trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 100.0,
  "eval_steps": 500,
  "global_step": 10600,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 0.9528496861457825,
      "learning_rate": 4.9500000000000004e-05,
      "loss": 1.0019,
      "step": 106
    },
    {
      "epoch": 1.0,
      "eval_LOCATION_f1": 0.0,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.0,
      "eval_LOCATION_recall": 0.0,
      "eval_ORGANIZATION_f1": 0.0,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.0,
      "eval_ORGANIZATION_recall": 0.0,
      "eval_PERSON_f1": 0.13658536585365852,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.2692307692307692,
      "eval_PERSON_recall": 0.0915032679738562,
      "eval_QUANTITY_f1": 0.0,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.0,
      "eval_QUANTITY_recall": 0.0,
      "eval_TIME_f1": 0.06060606060606061,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.5,
      "eval_TIME_recall": 0.03225806451612903,
      "eval_loss": 0.5637884140014648,
      "eval_overall_accuracy": 0.8441912311122121,
      "eval_overall_f1": 0.07246376811594202,
      "eval_overall_precision": 0.22058823529411764,
      "eval_overall_recall": 0.04335260115606936,
      "eval_runtime": 0.3279,
      "eval_samples_per_second": 570.368,
      "eval_steps_per_second": 9.15,
      "step": 106
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.2631016969680786,
      "learning_rate": 4.9e-05,
      "loss": 0.473,
      "step": 212
    },
    {
      "epoch": 2.0,
      "eval_LOCATION_f1": 0.2608695652173913,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.5217391304347826,
      "eval_LOCATION_recall": 0.17391304347826086,
      "eval_ORGANIZATION_f1": 0.13333333333333333,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.20689655172413793,
      "eval_ORGANIZATION_recall": 0.09836065573770492,
      "eval_PERSON_f1": 0.65564738292011,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.5666666666666667,
      "eval_PERSON_recall": 0.7777777777777778,
      "eval_QUANTITY_f1": 0.11320754716981132,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.14285714285714285,
      "eval_QUANTITY_recall": 0.09375,
      "eval_TIME_f1": 0.6233766233766235,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.5217391304347826,
      "eval_TIME_recall": 0.7741935483870968,
      "eval_loss": 0.32234013080596924,
      "eval_overall_accuracy": 0.9041367351994055,
      "eval_overall_f1": 0.48592592592592593,
      "eval_overall_precision": 0.49848024316109424,
      "eval_overall_recall": 0.47398843930635837,
      "eval_runtime": 0.3251,
      "eval_samples_per_second": 575.237,
      "eval_steps_per_second": 9.228,
      "step": 212
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.6863726377487183,
      "learning_rate": 4.85e-05,
      "loss": 0.3039,
      "step": 318
    },
    {
      "epoch": 3.0,
      "eval_LOCATION_f1": 0.5732484076433121,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.5113636363636364,
      "eval_LOCATION_recall": 0.6521739130434783,
      "eval_ORGANIZATION_f1": 0.46913580246913583,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.37623762376237624,
      "eval_ORGANIZATION_recall": 0.6229508196721312,
      "eval_PERSON_f1": 0.8217522658610271,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.7640449438202247,
      "eval_PERSON_recall": 0.8888888888888888,
      "eval_QUANTITY_f1": 0.6478873239436619,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.5897435897435898,
      "eval_QUANTITY_recall": 0.71875,
      "eval_TIME_f1": 0.7301587301587302,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.71875,
      "eval_TIME_recall": 0.7419354838709677,
      "eval_loss": 0.2088707536458969,
      "eval_overall_accuracy": 0.9348526133267278,
      "eval_overall_f1": 0.6760204081632653,
      "eval_overall_precision": 0.6050228310502284,
      "eval_overall_recall": 0.7658959537572254,
      "eval_runtime": 0.3264,
      "eval_samples_per_second": 572.929,
      "eval_steps_per_second": 9.191,
      "step": 318
    },
    {
      "epoch": 4.0,
      "grad_norm": 1.1164827346801758,
      "learning_rate": 4.8e-05,
      "loss": 0.2155,
      "step": 424
    },
    {
      "epoch": 4.0,
      "eval_LOCATION_f1": 0.6923076923076923,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.6206896551724138,
      "eval_LOCATION_recall": 0.782608695652174,
      "eval_ORGANIZATION_f1": 0.5853658536585367,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.46601941747572817,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.8553846153846153,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8081395348837209,
      "eval_PERSON_recall": 0.9084967320261438,
      "eval_QUANTITY_f1": 0.611764705882353,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.49056603773584906,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.6933333333333334,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.5909090909090909,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.19631721079349518,
      "eval_overall_accuracy": 0.9343571959375774,
      "eval_overall_f1": 0.7279503105590063,
      "eval_overall_precision": 0.6383442265795207,
      "eval_overall_recall": 0.846820809248555,
      "eval_runtime": 0.3268,
      "eval_samples_per_second": 572.217,
      "eval_steps_per_second": 9.18,
      "step": 424
    },
    {
      "epoch": 5.0,
      "grad_norm": 1.856414794921875,
      "learning_rate": 4.75e-05,
      "loss": 0.1851,
      "step": 530
    },
    {
      "epoch": 5.0,
      "eval_LOCATION_f1": 0.7382550335570469,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.6875,
      "eval_LOCATION_recall": 0.7971014492753623,
      "eval_ORGANIZATION_f1": 0.5314685314685316,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.4634146341463415,
      "eval_ORGANIZATION_recall": 0.6229508196721312,
      "eval_PERSON_f1": 0.8746177370030581,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8218390804597702,
      "eval_PERSON_recall": 0.934640522875817,
      "eval_QUANTITY_f1": 0.6944444444444444,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.625,
      "eval_QUANTITY_recall": 0.78125,
      "eval_TIME_f1": 0.8656716417910448,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8055555555555556,
      "eval_TIME_recall": 0.9354838709677419,
      "eval_loss": 0.14909601211547852,
      "eval_overall_accuracy": 0.9492197176120882,
      "eval_overall_f1": 0.7651715039577837,
      "eval_overall_precision": 0.7038834951456311,
      "eval_overall_recall": 0.838150289017341,
      "eval_runtime": 0.3263,
      "eval_samples_per_second": 573.095,
      "eval_steps_per_second": 9.194,
      "step": 530
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.673306405544281,
      "learning_rate": 4.7e-05,
      "loss": 0.1645,
      "step": 636
    },
    {
      "epoch": 6.0,
      "eval_LOCATION_f1": 0.7374999999999999,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.6483516483516484,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6482758620689655,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5595238095238095,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.8807339449541284,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8275862068965517,
      "eval_PERSON_recall": 0.9411764705882353,
      "eval_QUANTITY_f1": 0.7222222222222223,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.65,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.72,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.6136363636363636,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14334674179553986,
      "eval_overall_accuracy": 0.9524399306415655,
      "eval_overall_f1": 0.7779204107830552,
      "eval_overall_precision": 0.6997690531177829,
      "eval_overall_recall": 0.8757225433526011,
      "eval_runtime": 0.3274,
      "eval_samples_per_second": 571.25,
      "eval_steps_per_second": 9.164,
      "step": 636
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.47643959522247314,
      "learning_rate": 4.6500000000000005e-05,
      "loss": 0.1527,
      "step": 742
    },
    {
      "epoch": 7.0,
      "eval_LOCATION_f1": 0.7105263157894737,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.6506024096385542,
      "eval_LOCATION_recall": 0.782608695652174,
      "eval_ORGANIZATION_f1": 0.6535947712418301,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5434782608695652,
      "eval_ORGANIZATION_recall": 0.819672131147541,
      "eval_PERSON_f1": 0.8765432098765433,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8304093567251462,
      "eval_PERSON_recall": 0.9281045751633987,
      "eval_QUANTITY_f1": 0.767123287671233,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6829268292682927,
      "eval_QUANTITY_recall": 0.875,
      "eval_TIME_f1": 0.9032258064516129,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.9032258064516129,
      "eval_TIME_recall": 0.9032258064516129,
      "eval_loss": 0.13746976852416992,
      "eval_overall_accuracy": 0.9541738915035918,
      "eval_overall_f1": 0.7905759162303665,
      "eval_overall_precision": 0.722488038277512,
      "eval_overall_recall": 0.8728323699421965,
      "eval_runtime": 0.3275,
      "eval_samples_per_second": 570.917,
      "eval_steps_per_second": 9.159,
      "step": 742
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.47484996914863586,
      "learning_rate": 4.600000000000001e-05,
      "loss": 0.1415,
      "step": 848
    },
    {
      "epoch": 8.0,
      "eval_LOCATION_f1": 0.7346938775510203,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.6923076923076923,
      "eval_LOCATION_recall": 0.782608695652174,
      "eval_ORGANIZATION_f1": 0.6428571428571429,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.569620253164557,
      "eval_ORGANIZATION_recall": 0.7377049180327869,
      "eval_PERSON_f1": 0.8773006134969324,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8265895953757225,
      "eval_PERSON_recall": 0.934640522875817,
      "eval_QUANTITY_f1": 0.7323943661971831,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6666666666666666,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.7397260273972603,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.6428571428571429,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.13188910484313965,
      "eval_overall_accuracy": 0.9539261828090166,
      "eval_overall_f1": 0.7793923381770145,
      "eval_overall_precision": 0.7177615571776156,
      "eval_overall_recall": 0.8526011560693642,
      "eval_runtime": 0.3271,
      "eval_samples_per_second": 571.713,
      "eval_steps_per_second": 9.172,
      "step": 848
    },
    {
      "epoch": 9.0,
      "grad_norm": 1.1174803972244263,
      "learning_rate": 4.55e-05,
      "loss": 0.1336,
      "step": 954
    },
    {
      "epoch": 9.0,
      "eval_LOCATION_f1": 0.76,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7037037037037037,
      "eval_LOCATION_recall": 0.8260869565217391,
      "eval_ORGANIZATION_f1": 0.625,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5970149253731343,
      "eval_ORGANIZATION_recall": 0.6557377049180327,
      "eval_PERSON_f1": 0.88125,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.844311377245509,
      "eval_PERSON_recall": 0.9215686274509803,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.7761194029850746,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7222222222222222,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.12746429443359375,
      "eval_overall_accuracy": 0.9578895219222194,
      "eval_overall_f1": 0.7939972714870396,
      "eval_overall_precision": 0.751937984496124,
      "eval_overall_recall": 0.8410404624277457,
      "eval_runtime": 0.3268,
      "eval_samples_per_second": 572.154,
      "eval_steps_per_second": 9.179,
      "step": 954
    },
    {
      "epoch": 10.0,
      "grad_norm": 1.5846831798553467,
      "learning_rate": 4.5e-05,
      "loss": 0.1273,
      "step": 1060
    },
    {
      "epoch": 10.0,
      "eval_LOCATION_f1": 0.7832167832167832,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7567567567567568,
      "eval_LOCATION_recall": 0.8115942028985508,
      "eval_ORGANIZATION_f1": 0.6412213740458015,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6,
      "eval_ORGANIZATION_recall": 0.6885245901639344,
      "eval_PERSON_f1": 0.8923076923076924,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8430232558139535,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.12643250823020935,
      "eval_overall_accuracy": 0.9596234827842457,
      "eval_overall_f1": 0.8103683492496588,
      "eval_overall_precision": 0.7674418604651163,
      "eval_overall_recall": 0.8583815028901735,
      "eval_runtime": 0.3277,
      "eval_samples_per_second": 570.718,
      "eval_steps_per_second": 9.156,
      "step": 1060
    },
    {
      "epoch": 11.0,
      "grad_norm": 0.7376188635826111,
      "learning_rate": 4.4500000000000004e-05,
      "loss": 0.1208,
      "step": 1166
    },
    {
      "epoch": 11.0,
      "eval_LOCATION_f1": 0.7919463087248322,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7375,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6853146853146853,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5975609756097561,
      "eval_ORGANIZATION_recall": 0.8032786885245902,
      "eval_PERSON_f1": 0.9090909090909091,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8734939759036144,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.684931506849315,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6097560975609756,
      "eval_QUANTITY_recall": 0.78125,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.12870125472545624,
      "eval_overall_accuracy": 0.9568986871439188,
      "eval_overall_f1": 0.8133333333333334,
      "eval_overall_precision": 0.754950495049505,
      "eval_overall_recall": 0.8815028901734104,
      "eval_runtime": 0.3261,
      "eval_samples_per_second": 573.4,
      "eval_steps_per_second": 9.199,
      "step": 1166
    },
    {
      "epoch": 12.0,
      "grad_norm": 1.7949248552322388,
      "learning_rate": 4.4000000000000006e-05,
      "loss": 0.1152,
      "step": 1272
    },
    {
      "epoch": 12.0,
      "eval_LOCATION_f1": 0.7808219178082192,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7402597402597403,
      "eval_LOCATION_recall": 0.8260869565217391,
      "eval_ORGANIZATION_f1": 0.6818181818181818,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6338028169014085,
      "eval_ORGANIZATION_recall": 0.7377049180327869,
      "eval_PERSON_f1": 0.8902821316614421,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8554216867469879,
      "eval_PERSON_recall": 0.9281045751633987,
      "eval_QUANTITY_f1": 0.6933333333333334,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6046511627906976,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.835820895522388,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7777777777777778,
      "eval_TIME_recall": 0.9032258064516129,
      "eval_loss": 0.1358962059020996,
      "eval_overall_accuracy": 0.9564032697547684,
      "eval_overall_f1": 0.8064952638700946,
      "eval_overall_precision": 0.7582697201017812,
      "eval_overall_recall": 0.861271676300578,
      "eval_runtime": 0.3282,
      "eval_samples_per_second": 569.82,
      "eval_steps_per_second": 9.141,
      "step": 1272
    },
    {
      "epoch": 13.0,
      "grad_norm": 0.5476229190826416,
      "learning_rate": 4.35e-05,
      "loss": 0.1097,
      "step": 1378
    },
    {
      "epoch": 13.0,
      "eval_LOCATION_f1": 0.778523489932886,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.725,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.726027397260274,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6235294117647059,
      "eval_ORGANIZATION_recall": 0.8688524590163934,
      "eval_PERSON_f1": 0.9012345679012346,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8538011695906432,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7536231884057971,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7027027027027027,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.7941176470588235,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7297297297297297,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.130745992064476,
      "eval_overall_accuracy": 0.9593757740896706,
      "eval_overall_f1": 0.8201058201058201,
      "eval_overall_precision": 0.7560975609756098,
      "eval_overall_recall": 0.8959537572254336,
      "eval_runtime": 0.3271,
      "eval_samples_per_second": 571.737,
      "eval_steps_per_second": 9.172,
      "step": 1378
    },
    {
      "epoch": 14.0,
      "grad_norm": 1.329078197479248,
      "learning_rate": 4.3e-05,
      "loss": 0.1062,
      "step": 1484
    },
    {
      "epoch": 14.0,
      "eval_LOCATION_f1": 0.7919463087248322,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7375,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.697986577181208,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5909090909090909,
      "eval_ORGANIZATION_recall": 0.8524590163934426,
      "eval_PERSON_f1": 0.8978328173374613,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8529411764705882,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.72,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.627906976744186,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.7536231884057972,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.6842105263157895,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.14681027829647064,
      "eval_overall_accuracy": 0.9526876393361406,
      "eval_overall_f1": 0.8078431372549019,
      "eval_overall_precision": 0.7374701670644391,
      "eval_overall_recall": 0.8930635838150289,
      "eval_runtime": 0.327,
      "eval_samples_per_second": 571.95,
      "eval_steps_per_second": 9.176,
      "step": 1484
    },
    {
      "epoch": 15.0,
      "grad_norm": 0.9235780835151672,
      "learning_rate": 4.25e-05,
      "loss": 0.1034,
      "step": 1590
    },
    {
      "epoch": 15.0,
      "eval_LOCATION_f1": 0.7972972972972974,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7468354430379747,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.704225352112676,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6172839506172839,
      "eval_ORGANIZATION_recall": 0.819672131147541,
      "eval_PERSON_f1": 0.9050632911392406,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8773006134969326,
      "eval_PERSON_recall": 0.934640522875817,
      "eval_QUANTITY_f1": 0.7605633802816902,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6923076923076923,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.7536231884057972,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.6842105263157895,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.12539251148700714,
      "eval_overall_accuracy": 0.9568986871439188,
      "eval_overall_f1": 0.8176943699731902,
      "eval_overall_precision": 0.7625,
      "eval_overall_recall": 0.8815028901734104,
      "eval_runtime": 0.3262,
      "eval_samples_per_second": 573.286,
      "eval_steps_per_second": 9.197,
      "step": 1590
    },
    {
      "epoch": 16.0,
      "grad_norm": 0.6998082399368286,
      "learning_rate": 4.2e-05,
      "loss": 0.1019,
      "step": 1696
    },
    {
      "epoch": 16.0,
      "eval_LOCATION_f1": 0.7682119205298014,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7073170731707317,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.7210884353741497,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6162790697674418,
      "eval_ORGANIZATION_recall": 0.8688524590163934,
      "eval_PERSON_f1": 0.90625,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8682634730538922,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.7826086956521738,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7297297297297297,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8484848484848486,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8,
      "eval_TIME_recall": 0.9032258064516129,
      "eval_loss": 0.13638895750045776,
      "eval_overall_accuracy": 0.9576418132276443,
      "eval_overall_f1": 0.8260292164674634,
      "eval_overall_precision": 0.7641277641277642,
      "eval_overall_recall": 0.8988439306358381,
      "eval_runtime": 0.3263,
      "eval_samples_per_second": 573.085,
      "eval_steps_per_second": 9.194,
      "step": 1696
    },
    {
      "epoch": 17.0,
      "grad_norm": 0.25328242778778076,
      "learning_rate": 4.15e-05,
      "loss": 0.0961,
      "step": 1802
    },
    {
      "epoch": 17.0,
      "eval_LOCATION_f1": 0.7945205479452054,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7532467532467533,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.7034482758620689,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6071428571428571,
      "eval_ORGANIZATION_recall": 0.8360655737704918,
      "eval_PERSON_f1": 0.9056603773584905,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8727272727272727,
      "eval_PERSON_recall": 0.9411764705882353,
      "eval_QUANTITY_f1": 0.7222222222222223,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.65,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.7647058823529411,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7027027027027027,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.1412588506937027,
      "eval_overall_accuracy": 0.9546693088927422,
      "eval_overall_f1": 0.814419225634179,
      "eval_overall_precision": 0.7568238213399504,
      "eval_overall_recall": 0.8815028901734104,
      "eval_runtime": 0.3279,
      "eval_samples_per_second": 570.343,
      "eval_steps_per_second": 9.15,
      "step": 1802
    },
    {
      "epoch": 18.0,
      "grad_norm": 1.1196354627609253,
      "learning_rate": 4.1e-05,
      "loss": 0.0949,
      "step": 1908
    },
    {
      "epoch": 18.0,
      "eval_LOCATION_f1": 0.802721088435374,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7564102564102564,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6229508196721312,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6229508196721312,
      "eval_ORGANIZATION_recall": 0.6229508196721312,
      "eval_PERSON_f1": 0.9148264984227129,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8841463414634146,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.7826086956521738,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7297297297297297,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.7761194029850746,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7222222222222222,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.12308163195848465,
      "eval_overall_accuracy": 0.9603666088679713,
      "eval_overall_f1": 0.817174515235457,
      "eval_overall_precision": 0.7845744680851063,
      "eval_overall_recall": 0.8526011560693642,
      "eval_runtime": 0.326,
      "eval_samples_per_second": 573.624,
      "eval_steps_per_second": 9.203,
      "step": 1908
    },
    {
      "epoch": 19.0,
      "grad_norm": 1.150700569152832,
      "learning_rate": 4.05e-05,
      "loss": 0.09,
      "step": 2014
    },
    {
      "epoch": 19.0,
      "eval_LOCATION_f1": 0.7972972972972974,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7468354430379747,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6950354609929078,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6125,
      "eval_ORGANIZATION_recall": 0.8032786885245902,
      "eval_PERSON_f1": 0.9028213166144202,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8674698795180723,
      "eval_PERSON_recall": 0.9411764705882353,
      "eval_QUANTITY_f1": 0.7605633802816902,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6923076923076923,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.875,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8484848484848485,
      "eval_TIME_recall": 0.9032258064516129,
      "eval_loss": 0.12218116968870163,
      "eval_overall_accuracy": 0.9628436958137231,
      "eval_overall_f1": 0.8263795423956932,
      "eval_overall_precision": 0.7732997481108312,
      "eval_overall_recall": 0.8872832369942196,
      "eval_runtime": 0.3275,
      "eval_samples_per_second": 571.011,
      "eval_steps_per_second": 9.161,
      "step": 2014
    },
    {
      "epoch": 20.0,
      "grad_norm": 0.5211184024810791,
      "learning_rate": 4e-05,
      "loss": 0.0891,
      "step": 2120
    },
    {
      "epoch": 20.0,
      "eval_LOCATION_f1": 0.7945205479452054,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7532467532467533,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.7,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.620253164556962,
      "eval_ORGANIZATION_recall": 0.8032786885245902,
      "eval_PERSON_f1": 0.897196261682243,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8571428571428571,
      "eval_PERSON_recall": 0.9411764705882353,
      "eval_QUANTITY_f1": 0.7647058823529411,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7222222222222222,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.8125,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7878787878787878,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.13657507300376892,
      "eval_overall_accuracy": 0.9581372306167947,
      "eval_overall_f1": 0.8200270635994588,
      "eval_overall_precision": 0.7709923664122137,
      "eval_overall_recall": 0.8757225433526011,
      "eval_runtime": 0.3273,
      "eval_samples_per_second": 571.283,
      "eval_steps_per_second": 9.165,
      "step": 2120
    },
    {
      "epoch": 21.0,
      "grad_norm": 0.4734579920768738,
      "learning_rate": 3.9500000000000005e-05,
      "loss": 0.0865,
      "step": 2226
    },
    {
      "epoch": 21.0,
      "eval_LOCATION_f1": 0.7450980392156864,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.6785714285714286,
      "eval_LOCATION_recall": 0.8260869565217391,
      "eval_ORGANIZATION_f1": 0.6666666666666667,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5697674418604651,
      "eval_ORGANIZATION_recall": 0.8032786885245902,
      "eval_PERSON_f1": 0.9040247678018575,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8588235294117647,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7027027027027026,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6190476190476191,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.7647058823529411,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7027027027027027,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.14930059015750885,
      "eval_overall_accuracy": 0.9502105523903889,
      "eval_overall_f1": 0.7947712418300653,
      "eval_overall_precision": 0.7255369928400954,
      "eval_overall_recall": 0.8786127167630058,
      "eval_runtime": 0.3253,
      "eval_samples_per_second": 574.904,
      "eval_steps_per_second": 9.223,
      "step": 2226
    },
    {
      "epoch": 22.0,
      "grad_norm": 0.7941380739212036,
      "learning_rate": 3.9000000000000006e-05,
      "loss": 0.0889,
      "step": 2332
    },
    {
      "epoch": 22.0,
      "eval_LOCATION_f1": 0.7702702702702703,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7215189873417721,
      "eval_LOCATION_recall": 0.8260869565217391,
      "eval_ORGANIZATION_f1": 0.654320987654321,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5247524752475248,
      "eval_ORGANIZATION_recall": 0.8688524590163934,
      "eval_PERSON_f1": 0.896551724137931,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8614457831325302,
      "eval_PERSON_recall": 0.934640522875817,
      "eval_QUANTITY_f1": 0.75,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.675,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.7761194029850746,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7222222222222222,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.14907509088516235,
      "eval_overall_accuracy": 0.9494674263066634,
      "eval_overall_f1": 0.796875,
      "eval_overall_precision": 0.7251184834123223,
      "eval_overall_recall": 0.884393063583815,
      "eval_runtime": 0.3286,
      "eval_samples_per_second": 568.997,
      "eval_steps_per_second": 9.128,
      "step": 2332
    },
    {
      "epoch": 23.0,
      "grad_norm": 0.90423184633255,
      "learning_rate": 3.85e-05,
      "loss": 0.0847,
      "step": 2438
    },
    {
      "epoch": 23.0,
      "eval_LOCATION_f1": 0.7891156462585034,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7435897435897436,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.7205882352941175,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6533333333333333,
      "eval_ORGANIZATION_recall": 0.8032786885245902,
      "eval_PERSON_f1": 0.9028213166144202,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8674698795180723,
      "eval_PERSON_recall": 0.9411764705882353,
      "eval_QUANTITY_f1": 0.7222222222222223,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.65,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.7761194029850746,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7222222222222222,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.1280784159898758,
      "eval_overall_accuracy": 0.9591280653950953,
      "eval_overall_f1": 0.8178137651821862,
      "eval_overall_precision": 0.7670886075949367,
      "eval_overall_recall": 0.8757225433526011,
      "eval_runtime": 0.3258,
      "eval_samples_per_second": 573.892,
      "eval_steps_per_second": 9.207,
      "step": 2438
    },
    {
      "epoch": 24.0,
      "grad_norm": 1.168771505355835,
      "learning_rate": 3.8e-05,
      "loss": 0.0824,
      "step": 2544
    },
    {
      "epoch": 24.0,
      "eval_LOCATION_f1": 0.7945205479452054,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7532467532467533,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.6993006993006993,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6097560975609756,
      "eval_ORGANIZATION_recall": 0.819672131147541,
      "eval_PERSON_f1": 0.9006211180124225,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8579881656804734,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.7222222222222223,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.65,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.7887323943661972,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7,
      "eval_TIME_recall": 0.9032258064516129,
      "eval_loss": 0.14095881581306458,
      "eval_overall_accuracy": 0.9573941045330691,
      "eval_overall_f1": 0.8143236074270557,
      "eval_overall_precision": 0.7524509803921569,
      "eval_overall_recall": 0.8872832369942196,
      "eval_runtime": 0.327,
      "eval_samples_per_second": 571.879,
      "eval_steps_per_second": 9.175,
      "step": 2544
    },
    {
      "epoch": 25.0,
      "grad_norm": 0.6144793629646301,
      "learning_rate": 3.7500000000000003e-05,
      "loss": 0.0829,
      "step": 2650
    },
    {
      "epoch": 25.0,
      "eval_LOCATION_f1": 0.7702702702702703,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7215189873417721,
      "eval_LOCATION_recall": 0.8260869565217391,
      "eval_ORGANIZATION_f1": 0.6666666666666666,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5783132530120482,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9090909090909091,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8734939759036144,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.7714285714285714,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7105263157894737,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.823529411764706,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7567567567567568,
      "eval_TIME_recall": 0.9032258064516129,
      "eval_loss": 0.13440415263175964,
      "eval_overall_accuracy": 0.9593757740896706,
      "eval_overall_f1": 0.814419225634179,
      "eval_overall_precision": 0.7568238213399504,
      "eval_overall_recall": 0.8815028901734104,
      "eval_runtime": 0.3261,
      "eval_samples_per_second": 573.389,
      "eval_steps_per_second": 9.199,
      "step": 2650
    },
    {
      "epoch": 26.0,
      "grad_norm": 0.7332161068916321,
      "learning_rate": 3.7e-05,
      "loss": 0.0786,
      "step": 2756
    },
    {
      "epoch": 26.0,
      "eval_LOCATION_f1": 0.7972972972972974,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7468354430379747,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6956521739130435,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6233766233766234,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9056603773584905,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8727272727272727,
      "eval_PERSON_recall": 0.9411764705882353,
      "eval_QUANTITY_f1": 0.7605633802816902,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6923076923076923,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.787878787878788,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7428571428571429,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.1297662854194641,
      "eval_overall_accuracy": 0.9603666088679713,
      "eval_overall_f1": 0.8205128205128205,
      "eval_overall_precision": 0.769620253164557,
      "eval_overall_recall": 0.8786127167630058,
      "eval_runtime": 0.3277,
      "eval_samples_per_second": 570.619,
      "eval_steps_per_second": 9.154,
      "step": 2756
    },
    {
      "epoch": 27.0,
      "grad_norm": 0.9926333427429199,
      "learning_rate": 3.65e-05,
      "loss": 0.0786,
      "step": 2862
    },
    {
      "epoch": 27.0,
      "eval_LOCATION_f1": 0.7837837837837838,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7341772151898734,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.6911764705882353,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6266666666666667,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.90625,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8682634730538922,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.7605633802816902,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6923076923076923,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.7647058823529411,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7027027027027027,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.13430674374103546,
      "eval_overall_accuracy": 0.961357443646272,
      "eval_overall_f1": 0.8156123822341858,
      "eval_overall_precision": 0.7632241813602015,
      "eval_overall_recall": 0.8757225433526011,
      "eval_runtime": 0.3263,
      "eval_samples_per_second": 573.041,
      "eval_steps_per_second": 9.193,
      "step": 2862
    },
    {
      "epoch": 28.0,
      "grad_norm": 1.1402696371078491,
      "learning_rate": 3.6e-05,
      "loss": 0.0753,
      "step": 2968
    },
    {
      "epoch": 28.0,
      "eval_LOCATION_f1": 0.7814569536423841,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6461538461538463,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6086956521739131,
      "eval_ORGANIZATION_recall": 0.6885245901639344,
      "eval_PERSON_f1": 0.90625,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8682634730538922,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.7826086956521738,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7297297297297297,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.7323943661971831,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.65,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.12820479273796082,
      "eval_overall_accuracy": 0.9598711914788209,
      "eval_overall_f1": 0.8070175438596492,
      "eval_overall_precision": 0.7569620253164557,
      "eval_overall_recall": 0.8641618497109826,
      "eval_runtime": 0.3275,
      "eval_samples_per_second": 571.063,
      "eval_steps_per_second": 9.161,
      "step": 2968
    },
    {
      "epoch": 29.0,
      "grad_norm": 7.802056789398193,
      "learning_rate": 3.55e-05,
      "loss": 0.0736,
      "step": 3074
    },
    {
      "epoch": 29.0,
      "eval_LOCATION_f1": 0.8137931034482757,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7763157894736842,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6388888888888888,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5542168674698795,
      "eval_ORGANIZATION_recall": 0.7540983606557377,
      "eval_PERSON_f1": 0.9119496855345912,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8787878787878788,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8253968253968254,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8125,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.13266420364379883,
      "eval_overall_accuracy": 0.9601189001733961,
      "eval_overall_f1": 0.8211382113821138,
      "eval_overall_precision": 0.7729591836734694,
      "eval_overall_recall": 0.8757225433526011,
      "eval_runtime": 0.3254,
      "eval_samples_per_second": 574.742,
      "eval_steps_per_second": 9.22,
      "step": 3074
    },
    {
      "epoch": 30.0,
      "grad_norm": 0.6352629661560059,
      "learning_rate": 3.5e-05,
      "loss": 0.0718,
      "step": 3180
    },
    {
      "epoch": 30.0,
      "eval_LOCATION_f1": 0.8082191780821918,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7662337662337663,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6666666666666667,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5697674418604651,
      "eval_ORGANIZATION_recall": 0.8032786885245902,
      "eval_PERSON_f1": 0.9125,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.874251497005988,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.7761194029850746,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7222222222222222,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.13187818229198456,
      "eval_overall_accuracy": 0.9601189001733961,
      "eval_overall_f1": 0.8208556149732621,
      "eval_overall_precision": 0.763681592039801,
      "eval_overall_recall": 0.8872832369942196,
      "eval_runtime": 0.3252,
      "eval_samples_per_second": 575.055,
      "eval_steps_per_second": 9.225,
      "step": 3180
    },
    {
      "epoch": 31.0,
      "grad_norm": 0.4606497287750244,
      "learning_rate": 3.45e-05,
      "loss": 0.069,
      "step": 3286
    },
    {
      "epoch": 31.0,
      "eval_LOCATION_f1": 0.7651006711409397,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7125,
      "eval_LOCATION_recall": 0.8260869565217391,
      "eval_ORGANIZATION_f1": 0.6564885496183206,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6142857142857143,
      "eval_ORGANIZATION_recall": 0.7049180327868853,
      "eval_PERSON_f1": 0.9068322981366459,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.863905325443787,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7246376811594203,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6756756756756757,
      "eval_QUANTITY_recall": 0.78125,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.717948717948718,
      "eval_TIME_recall": 0.9032258064516129,
      "eval_loss": 0.13368181884288788,
      "eval_overall_accuracy": 0.9621005697299975,
      "eval_overall_f1": 0.8070175438596492,
      "eval_overall_precision": 0.7569620253164557,
      "eval_overall_recall": 0.8641618497109826,
      "eval_runtime": 0.3258,
      "eval_samples_per_second": 573.934,
      "eval_steps_per_second": 9.207,
      "step": 3286
    },
    {
      "epoch": 32.0,
      "grad_norm": 0.6008712649345398,
      "learning_rate": 3.4000000000000007e-05,
      "loss": 0.0688,
      "step": 3392
    },
    {
      "epoch": 32.0,
      "eval_LOCATION_f1": 0.7682119205298014,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7073170731707317,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.6865671641791045,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6301369863013698,
      "eval_ORGANIZATION_recall": 0.7540983606557377,
      "eval_PERSON_f1": 0.9090909090909091,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8734939759036144,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.7761194029850748,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7428571428571429,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.7647058823529411,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7027027027027027,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.12878015637397766,
      "eval_overall_accuracy": 0.961357443646272,
      "eval_overall_f1": 0.8146143437077131,
      "eval_overall_precision": 0.7659033078880407,
      "eval_overall_recall": 0.869942196531792,
      "eval_runtime": 0.3264,
      "eval_samples_per_second": 572.956,
      "eval_steps_per_second": 9.192,
      "step": 3392
    },
    {
      "epoch": 33.0,
      "grad_norm": 0.6221128106117249,
      "learning_rate": 3.35e-05,
      "loss": 0.0733,
      "step": 3498
    },
    {
      "epoch": 33.0,
      "eval_LOCATION_f1": 0.778523489932886,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.725,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.682170542635659,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6470588235294118,
      "eval_ORGANIZATION_recall": 0.7213114754098361,
      "eval_PERSON_f1": 0.9090909090909091,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8734939759036144,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.7761194029850746,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7222222222222222,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.12517035007476807,
      "eval_overall_accuracy": 0.9643299479811741,
      "eval_overall_f1": 0.8196721311475409,
      "eval_overall_precision": 0.7772020725388601,
      "eval_overall_recall": 0.8670520231213873,
      "eval_runtime": 0.3272,
      "eval_samples_per_second": 571.566,
      "eval_steps_per_second": 9.17,
      "step": 3498
    },
    {
      "epoch": 34.0,
      "grad_norm": 0.5183610916137695,
      "learning_rate": 3.3e-05,
      "loss": 0.0681,
      "step": 3604
    },
    {
      "epoch": 34.0,
      "eval_LOCATION_f1": 0.7682119205298014,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7073170731707317,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.6857142857142857,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6075949367088608,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.897196261682243,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8571428571428571,
      "eval_PERSON_recall": 0.9411764705882353,
      "eval_QUANTITY_f1": 0.7714285714285714,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7105263157894737,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.782608695652174,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7105263157894737,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.13874541223049164,
      "eval_overall_accuracy": 0.9601189001733961,
      "eval_overall_f1": 0.8095872170439414,
      "eval_overall_precision": 0.7506172839506173,
      "eval_overall_recall": 0.8786127167630058,
      "eval_runtime": 0.3256,
      "eval_samples_per_second": 574.251,
      "eval_steps_per_second": 9.213,
      "step": 3604
    },
    {
      "epoch": 35.0,
      "grad_norm": 1.8626983165740967,
      "learning_rate": 3.2500000000000004e-05,
      "loss": 0.0683,
      "step": 3710
    },
    {
      "epoch": 35.0,
      "eval_LOCATION_f1": 0.763157894736842,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.6987951807228916,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.676056338028169,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5925925925925926,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9085173501577287,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8780487804878049,
      "eval_PERSON_recall": 0.9411764705882353,
      "eval_QUANTITY_f1": 0.7123287671232876,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6341463414634146,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.84375,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8181818181818182,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.13318687677383423,
      "eval_overall_accuracy": 0.9603666088679713,
      "eval_overall_f1": 0.8101604278074866,
      "eval_overall_precision": 0.753731343283582,
      "eval_overall_recall": 0.8757225433526011,
      "eval_runtime": 0.3274,
      "eval_samples_per_second": 571.196,
      "eval_steps_per_second": 9.164,
      "step": 3710
    },
    {
      "epoch": 36.0,
      "grad_norm": 0.7150887250900269,
      "learning_rate": 3.2000000000000005e-05,
      "loss": 0.0699,
      "step": 3816
    },
    {
      "epoch": 36.0,
      "eval_LOCATION_f1": 0.7947019867549668,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6620689655172414,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5714285714285714,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9113924050632911,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8834355828220859,
      "eval_PERSON_recall": 0.9411764705882353,
      "eval_QUANTITY_f1": 0.7536231884057971,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7027027027027027,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.84375,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8181818181818182,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.13158132135868073,
      "eval_overall_accuracy": 0.9596234827842457,
      "eval_overall_f1": 0.8187919463087249,
      "eval_overall_precision": 0.7644110275689223,
      "eval_overall_recall": 0.8815028901734104,
      "eval_runtime": 0.3263,
      "eval_samples_per_second": 573.064,
      "eval_steps_per_second": 9.194,
      "step": 3816
    },
    {
      "epoch": 37.0,
      "grad_norm": 1.7562237977981567,
      "learning_rate": 3.15e-05,
      "loss": 0.0648,
      "step": 3922
    },
    {
      "epoch": 37.0,
      "eval_LOCATION_f1": 0.8163265306122449,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7692307692307693,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6277372262773723,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5657894736842105,
      "eval_ORGANIZATION_recall": 0.7049180327868853,
      "eval_PERSON_f1": 0.9148264984227129,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8841463414634146,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.787878787878788,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7647058823529411,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.8571428571428571,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.84375,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.12392880767583847,
      "eval_overall_accuracy": 0.9633391132028735,
      "eval_overall_f1": 0.8246575342465753,
      "eval_overall_precision": 0.7838541666666666,
      "eval_overall_recall": 0.869942196531792,
      "eval_runtime": 0.3269,
      "eval_samples_per_second": 572.004,
      "eval_steps_per_second": 9.177,
      "step": 3922
    },
    {
      "epoch": 38.0,
      "grad_norm": 1.6564940214157104,
      "learning_rate": 3.1e-05,
      "loss": 0.0653,
      "step": 4028
    },
    {
      "epoch": 38.0,
      "eval_LOCATION_f1": 0.7651006711409397,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7125,
      "eval_LOCATION_recall": 0.8260869565217391,
      "eval_ORGANIZATION_f1": 0.6993006993006993,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6097560975609756,
      "eval_ORGANIZATION_recall": 0.819672131147541,
      "eval_PERSON_f1": 0.9119496855345912,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8787878787878788,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.7323943661971831,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6666666666666666,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.12974902987480164,
      "eval_overall_accuracy": 0.9628436958137231,
      "eval_overall_f1": 0.8176943699731902,
      "eval_overall_precision": 0.7625,
      "eval_overall_recall": 0.8815028901734104,
      "eval_runtime": 0.3255,
      "eval_samples_per_second": 574.439,
      "eval_steps_per_second": 9.216,
      "step": 4028
    },
    {
      "epoch": 39.0,
      "grad_norm": 0.9774243235588074,
      "learning_rate": 3.05e-05,
      "loss": 0.0627,
      "step": 4134
    },
    {
      "epoch": 39.0,
      "eval_LOCATION_f1": 0.7999999999999999,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7407407407407407,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6470588235294118,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5866666666666667,
      "eval_ORGANIZATION_recall": 0.7213114754098361,
      "eval_PERSON_f1": 0.9182389937106918,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8848484848484849,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7323943661971831,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6666666666666666,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.8059701492537312,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.75,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.127953439950943,
      "eval_overall_accuracy": 0.9625959871191478,
      "eval_overall_f1": 0.8167115902964959,
      "eval_overall_precision": 0.7651515151515151,
      "eval_overall_recall": 0.8757225433526011,
      "eval_runtime": 0.3268,
      "eval_samples_per_second": 572.268,
      "eval_steps_per_second": 9.181,
      "step": 4134
    },
    {
      "epoch": 40.0,
      "grad_norm": 1.1059598922729492,
      "learning_rate": 3e-05,
      "loss": 0.0636,
      "step": 4240
    },
    {
      "epoch": 40.0,
      "eval_LOCATION_f1": 0.802721088435374,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7564102564102564,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6176470588235294,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.56,
      "eval_ORGANIZATION_recall": 0.6885245901639344,
      "eval_PERSON_f1": 0.9216300940438872,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8855421686746988,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.75,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.675,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.13029487431049347,
      "eval_overall_accuracy": 0.9621005697299975,
      "eval_overall_f1": 0.8162162162162162,
      "eval_overall_precision": 0.766497461928934,
      "eval_overall_recall": 0.8728323699421965,
      "eval_runtime": 0.3257,
      "eval_samples_per_second": 574.23,
      "eval_steps_per_second": 9.212,
      "step": 4240
    },
    {
      "epoch": 41.0,
      "grad_norm": 0.2690298557281494,
      "learning_rate": 2.95e-05,
      "loss": 0.0611,
      "step": 4346
    },
    {
      "epoch": 41.0,
      "eval_LOCATION_f1": 0.7947019867549668,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6853146853146853,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5975609756097561,
      "eval_ORGANIZATION_recall": 0.8032786885245902,
      "eval_PERSON_f1": 0.927899686520376,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.891566265060241,
      "eval_PERSON_recall": 0.9673202614379085,
      "eval_QUANTITY_f1": 0.7605633802816902,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6923076923076923,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14014151692390442,
      "eval_overall_accuracy": 0.9603666088679713,
      "eval_overall_f1": 0.8304405874499332,
      "eval_overall_precision": 0.771712158808933,
      "eval_overall_recall": 0.8988439306358381,
      "eval_runtime": 0.326,
      "eval_samples_per_second": 573.661,
      "eval_steps_per_second": 9.203,
      "step": 4346
    },
    {
      "epoch": 42.0,
      "grad_norm": 0.555206835269928,
      "learning_rate": 2.9e-05,
      "loss": 0.0622,
      "step": 4452
    },
    {
      "epoch": 42.0,
      "eval_LOCATION_f1": 0.7919463087248322,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7375,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.7007299270072993,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.631578947368421,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9274447949526814,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8963414634146342,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7714285714285714,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7105263157894737,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14087490737438202,
      "eval_overall_accuracy": 0.9623482784245727,
      "eval_overall_f1": 0.8346883468834689,
      "eval_overall_precision": 0.7857142857142857,
      "eval_overall_recall": 0.8901734104046243,
      "eval_runtime": 0.3274,
      "eval_samples_per_second": 571.246,
      "eval_steps_per_second": 9.164,
      "step": 4452
    },
    {
      "epoch": 43.0,
      "grad_norm": 0.2941405475139618,
      "learning_rate": 2.8499999999999998e-05,
      "loss": 0.0572,
      "step": 4558
    },
    {
      "epoch": 43.0,
      "eval_LOCATION_f1": 0.7866666666666667,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.689655172413793,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5952380952380952,
      "eval_ORGANIZATION_recall": 0.819672131147541,
      "eval_PERSON_f1": 0.915360501567398,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8795180722891566,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7826086956521738,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7297297297297297,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15049412846565247,
      "eval_overall_accuracy": 0.9591280653950953,
      "eval_overall_f1": 0.8262032085561497,
      "eval_overall_precision": 0.7686567164179104,
      "eval_overall_recall": 0.8930635838150289,
      "eval_runtime": 0.3257,
      "eval_samples_per_second": 574.133,
      "eval_steps_per_second": 9.211,
      "step": 4558
    },
    {
      "epoch": 44.0,
      "grad_norm": 1.0385116338729858,
      "learning_rate": 2.8000000000000003e-05,
      "loss": 0.0574,
      "step": 4664
    },
    {
      "epoch": 44.0,
      "eval_LOCATION_f1": 0.7972972972972974,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7468354430379747,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.656934306569343,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5921052631578947,
      "eval_ORGANIZATION_recall": 0.7377049180327869,
      "eval_PERSON_f1": 0.909657320872274,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8690476190476191,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7536231884057971,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7027027027027027,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.13434460759162903,
      "eval_overall_accuracy": 0.9625959871191478,
      "eval_overall_f1": 0.8189189189189189,
      "eval_overall_precision": 0.7690355329949239,
      "eval_overall_recall": 0.8757225433526011,
      "eval_runtime": 0.3252,
      "eval_samples_per_second": 575.072,
      "eval_steps_per_second": 9.226,
      "step": 4664
    },
    {
      "epoch": 45.0,
      "grad_norm": 1.3497114181518555,
      "learning_rate": 2.7500000000000004e-05,
      "loss": 0.0602,
      "step": 4770
    },
    {
      "epoch": 45.0,
      "eval_LOCATION_f1": 0.8163265306122449,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7692307692307693,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6259541984732824,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5857142857142857,
      "eval_ORGANIZATION_recall": 0.6721311475409836,
      "eval_PERSON_f1": 0.915360501567398,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8795180722891566,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7605633802816902,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6923076923076923,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.13049767911434174,
      "eval_overall_accuracy": 0.961357443646272,
      "eval_overall_f1": 0.8201634877384196,
      "eval_overall_precision": 0.7757731958762887,
      "eval_overall_recall": 0.869942196531792,
      "eval_runtime": 0.3258,
      "eval_samples_per_second": 573.995,
      "eval_steps_per_second": 9.208,
      "step": 4770
    },
    {
      "epoch": 46.0,
      "grad_norm": 0.6758576035499573,
      "learning_rate": 2.7000000000000002e-05,
      "loss": 0.0571,
      "step": 4876
    },
    {
      "epoch": 46.0,
      "eval_LOCATION_f1": 0.8079470198675496,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8840579710144928,
      "eval_ORGANIZATION_f1": 0.6212121212121212,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5774647887323944,
      "eval_ORGANIZATION_recall": 0.6721311475409836,
      "eval_PERSON_f1": 0.9119496855345912,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8787878787878788,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.7826086956521738,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7297297297297297,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.13599485158920288,
      "eval_overall_accuracy": 0.9606143175625464,
      "eval_overall_f1": 0.819047619047619,
      "eval_overall_precision": 0.7737789203084833,
      "eval_overall_recall": 0.869942196531792,
      "eval_runtime": 0.3246,
      "eval_samples_per_second": 576.084,
      "eval_steps_per_second": 9.242,
      "step": 4876
    },
    {
      "epoch": 47.0,
      "grad_norm": 0.7021151781082153,
      "learning_rate": 2.6500000000000004e-05,
      "loss": 0.0566,
      "step": 4982
    },
    {
      "epoch": 47.0,
      "eval_LOCATION_f1": 0.8108108108108109,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6666666666666666,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5974025974025974,
      "eval_ORGANIZATION_recall": 0.7540983606557377,
      "eval_PERSON_f1": 0.9211356466876972,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8902439024390244,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7647058823529411,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7222222222222222,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.1346343755722046,
      "eval_overall_accuracy": 0.9630914045082982,
      "eval_overall_f1": 0.8276797829036635,
      "eval_overall_precision": 0.7800511508951407,
      "eval_overall_recall": 0.8815028901734104,
      "eval_runtime": 0.3248,
      "eval_samples_per_second": 575.823,
      "eval_steps_per_second": 9.238,
      "step": 4982
    },
    {
      "epoch": 48.0,
      "grad_norm": 1.4580228328704834,
      "learning_rate": 2.6000000000000002e-05,
      "loss": 0.0577,
      "step": 5088
    },
    {
      "epoch": 48.0,
      "eval_LOCATION_f1": 0.7999999999999999,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7407407407407407,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6376811594202898,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5714285714285714,
      "eval_ORGANIZATION_recall": 0.7213114754098361,
      "eval_PERSON_f1": 0.9119496855345912,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8787878787878788,
      "eval_PERSON_recall": 0.9477124183006536,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8571428571428571,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.84375,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.13921397924423218,
      "eval_overall_accuracy": 0.9633391132028735,
      "eval_overall_f1": 0.8222523744911805,
      "eval_overall_precision": 0.7749360613810742,
      "eval_overall_recall": 0.8757225433526011,
      "eval_runtime": 0.3257,
      "eval_samples_per_second": 574.082,
      "eval_steps_per_second": 9.21,
      "step": 5088
    },
    {
      "epoch": 49.0,
      "grad_norm": 1.2305538654327393,
      "learning_rate": 2.5500000000000003e-05,
      "loss": 0.0555,
      "step": 5194
    },
    {
      "epoch": 49.0,
      "eval_LOCATION_f1": 0.7947019867549668,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6865671641791045,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6301369863013698,
      "eval_ORGANIZATION_recall": 0.7540983606557377,
      "eval_PERSON_f1": 0.9269841269841269,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9012345679012346,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7536231884057971,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7027027027027027,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.84375,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8181818181818182,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.13334891200065613,
      "eval_overall_accuracy": 0.9653207827594749,
      "eval_overall_f1": 0.8321964529331515,
      "eval_overall_precision": 0.7881136950904393,
      "eval_overall_recall": 0.8815028901734104,
      "eval_runtime": 0.3272,
      "eval_samples_per_second": 571.531,
      "eval_steps_per_second": 9.169,
      "step": 5194
    },
    {
      "epoch": 50.0,
      "grad_norm": 1.1206848621368408,
      "learning_rate": 2.5e-05,
      "loss": 0.057,
      "step": 5300
    },
    {
      "epoch": 50.0,
      "eval_LOCATION_f1": 0.7763157894736842,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7108433734939759,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6176470588235294,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.56,
      "eval_ORGANIZATION_recall": 0.6885245901639344,
      "eval_PERSON_f1": 0.915360501567398,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8795180722891566,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7647058823529411,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7222222222222222,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.1372886449098587,
      "eval_overall_accuracy": 0.9606143175625464,
      "eval_overall_f1": 0.8108108108108107,
      "eval_overall_precision": 0.7614213197969543,
      "eval_overall_recall": 0.8670520231213873,
      "eval_runtime": 0.3261,
      "eval_samples_per_second": 573.416,
      "eval_steps_per_second": 9.199,
      "step": 5300
    },
    {
      "epoch": 51.0,
      "grad_norm": 0.7590587735176086,
      "learning_rate": 2.45e-05,
      "loss": 0.0539,
      "step": 5406
    },
    {
      "epoch": 51.0,
      "eval_LOCATION_f1": 0.7999999999999999,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7407407407407407,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6713286713286712,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5853658536585366,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9240506329113924,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8957055214723927,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7826086956521738,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7297297297297297,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.84375,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8181818181818182,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.1392890363931656,
      "eval_overall_accuracy": 0.964082239286599,
      "eval_overall_f1": 0.8301886792452831,
      "eval_overall_precision": 0.7777777777777778,
      "eval_overall_recall": 0.8901734104046243,
      "eval_runtime": 0.3266,
      "eval_samples_per_second": 572.617,
      "eval_steps_per_second": 9.186,
      "step": 5406
    },
    {
      "epoch": 52.0,
      "grad_norm": 0.46395185589790344,
      "learning_rate": 2.4e-05,
      "loss": 0.0528,
      "step": 5512
    },
    {
      "epoch": 52.0,
      "eval_LOCATION_f1": 0.8163265306122449,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7692307692307693,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.7132867132867133,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6219512195121951,
      "eval_ORGANIZATION_recall": 0.8360655737704918,
      "eval_PERSON_f1": 0.930817610062893,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.896969696969697,
      "eval_PERSON_recall": 0.9673202614379085,
      "eval_QUANTITY_f1": 0.8059701492537314,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7714285714285715,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14341352880001068,
      "eval_overall_accuracy": 0.9643299479811741,
      "eval_overall_f1": 0.844804318488529,
      "eval_overall_precision": 0.7924050632911392,
      "eval_overall_recall": 0.9046242774566474,
      "eval_runtime": 0.3254,
      "eval_samples_per_second": 574.743,
      "eval_steps_per_second": 9.22,
      "step": 5512
    },
    {
      "epoch": 53.0,
      "grad_norm": 1.4821114540100098,
      "learning_rate": 2.35e-05,
      "loss": 0.0539,
      "step": 5618
    },
    {
      "epoch": 53.0,
      "eval_LOCATION_f1": 0.8053691275167786,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.75,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.704225352112676,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6172839506172839,
      "eval_ORGANIZATION_recall": 0.819672131147541,
      "eval_PERSON_f1": 0.9187500000000001,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8802395209580839,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7714285714285714,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7105263157894737,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8571428571428571,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.84375,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.13907547295093536,
      "eval_overall_accuracy": 0.9645776566757494,
      "eval_overall_f1": 0.836021505376344,
      "eval_overall_precision": 0.7814070351758794,
      "eval_overall_recall": 0.8988439306358381,
      "eval_runtime": 0.3261,
      "eval_samples_per_second": 573.484,
      "eval_steps_per_second": 9.2,
      "step": 5618
    },
    {
      "epoch": 54.0,
      "grad_norm": 0.7181128859519958,
      "learning_rate": 2.3000000000000003e-05,
      "loss": 0.0514,
      "step": 5724
    },
    {
      "epoch": 54.0,
      "eval_LOCATION_f1": 0.7999999999999999,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7407407407407407,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6944444444444445,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6024096385542169,
      "eval_ORGANIZATION_recall": 0.819672131147541,
      "eval_PERSON_f1": 0.9216300940438872,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8855421686746988,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7826086956521738,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7297297297297297,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8125,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7878787878787878,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.14943979680538177,
      "eval_overall_accuracy": 0.9628436958137231,
      "eval_overall_f1": 0.8310991957104559,
      "eval_overall_precision": 0.775,
      "eval_overall_recall": 0.8959537572254336,
      "eval_runtime": 0.326,
      "eval_samples_per_second": 573.678,
      "eval_steps_per_second": 9.203,
      "step": 5724
    },
    {
      "epoch": 55.0,
      "grad_norm": 1.2386493682861328,
      "learning_rate": 2.25e-05,
      "loss": 0.0544,
      "step": 5830
    },
    {
      "epoch": 55.0,
      "eval_LOCATION_f1": 0.8133333333333332,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7530864197530864,
      "eval_LOCATION_recall": 0.8840579710144928,
      "eval_ORGANIZATION_f1": 0.7194244604316549,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6410256410256411,
      "eval_ORGANIZATION_recall": 0.819672131147541,
      "eval_PERSON_f1": 0.915360501567398,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8795180722891566,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7605633802816902,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6923076923076923,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8059701492537312,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.75,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14867006242275238,
      "eval_overall_accuracy": 0.9623482784245727,
      "eval_overall_f1": 0.8337801608579088,
      "eval_overall_precision": 0.7775,
      "eval_overall_recall": 0.8988439306358381,
      "eval_runtime": 0.3267,
      "eval_samples_per_second": 572.462,
      "eval_steps_per_second": 9.184,
      "step": 5830
    },
    {
      "epoch": 56.0,
      "grad_norm": 1.9313570261001587,
      "learning_rate": 2.2000000000000003e-05,
      "loss": 0.0508,
      "step": 5936
    },
    {
      "epoch": 56.0,
      "eval_LOCATION_f1": 0.7682119205298014,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7073170731707317,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.6911764705882353,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6266666666666667,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.9274447949526814,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8963414634146342,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.742857142857143,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.6842105263157895,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.84375,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8181818181818182,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.1412782073020935,
      "eval_overall_accuracy": 0.9628436958137231,
      "eval_overall_f1": 0.8265582655826559,
      "eval_overall_precision": 0.7780612244897959,
      "eval_overall_recall": 0.8815028901734104,
      "eval_runtime": 0.328,
      "eval_samples_per_second": 570.044,
      "eval_steps_per_second": 9.145,
      "step": 5936
    },
    {
      "epoch": 57.0,
      "grad_norm": 0.9018868207931519,
      "learning_rate": 2.15e-05,
      "loss": 0.0501,
      "step": 6042
    },
    {
      "epoch": 57.0,
      "eval_LOCATION_f1": 0.7814569536423841,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.7007299270072993,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.631578947368421,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9303797468354431,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.901840490797546,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.146066352725029,
      "eval_overall_accuracy": 0.964082239286599,
      "eval_overall_f1": 0.835820895522388,
      "eval_overall_precision": 0.7877237851662404,
      "eval_overall_recall": 0.8901734104046243,
      "eval_runtime": 0.3253,
      "eval_samples_per_second": 574.891,
      "eval_steps_per_second": 9.223,
      "step": 6042
    },
    {
      "epoch": 58.0,
      "grad_norm": 0.23339399695396423,
      "learning_rate": 2.1e-05,
      "loss": 0.05,
      "step": 6148
    },
    {
      "epoch": 58.0,
      "eval_LOCATION_f1": 0.7972972972972974,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7468354430379747,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6956521739130435,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6233766233766234,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9303797468354431,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.901840490797546,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7714285714285714,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7105263157894737,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14607301354408264,
      "eval_overall_accuracy": 0.9630914045082982,
      "eval_overall_f1": 0.835820895522388,
      "eval_overall_precision": 0.7877237851662404,
      "eval_overall_recall": 0.8901734104046243,
      "eval_runtime": 0.3264,
      "eval_samples_per_second": 572.942,
      "eval_steps_per_second": 9.192,
      "step": 6148
    },
    {
      "epoch": 59.0,
      "grad_norm": 0.6318117380142212,
      "learning_rate": 2.05e-05,
      "loss": 0.0487,
      "step": 6254
    },
    {
      "epoch": 59.0,
      "eval_LOCATION_f1": 0.7814569536423841,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.7101449275362318,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6363636363636364,
      "eval_ORGANIZATION_recall": 0.8032786885245902,
      "eval_PERSON_f1": 0.9274447949526814,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8963414634146342,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.8059701492537314,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7714285714285715,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.84375,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8181818181818182,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14298133552074432,
      "eval_overall_accuracy": 0.9648253653703245,
      "eval_overall_f1": 0.8385345997286295,
      "eval_overall_precision": 0.7902813299232737,
      "eval_overall_recall": 0.8930635838150289,
      "eval_runtime": 0.3261,
      "eval_samples_per_second": 573.528,
      "eval_steps_per_second": 9.201,
      "step": 6254
    },
    {
      "epoch": 60.0,
      "grad_norm": 0.3348062336444855,
      "learning_rate": 2e-05,
      "loss": 0.0479,
      "step": 6360
    },
    {
      "epoch": 60.0,
      "eval_LOCATION_f1": 0.778523489932886,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.725,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.7272727272727272,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6341463414634146,
      "eval_ORGANIZATION_recall": 0.8524590163934426,
      "eval_PERSON_f1": 0.9303797468354431,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.901840490797546,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7647058823529411,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7222222222222222,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.7941176470588235,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7297297297297297,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.1541959047317505,
      "eval_overall_accuracy": 0.9623482784245727,
      "eval_overall_f1": 0.8333333333333334,
      "eval_overall_precision": 0.7788944723618091,
      "eval_overall_recall": 0.8959537572254336,
      "eval_runtime": 0.3261,
      "eval_samples_per_second": 573.376,
      "eval_steps_per_second": 9.199,
      "step": 6360
    },
    {
      "epoch": 61.0,
      "grad_norm": 0.7797447443008423,
      "learning_rate": 1.9500000000000003e-05,
      "loss": 0.05,
      "step": 6466
    },
    {
      "epoch": 61.0,
      "eval_LOCATION_f1": 0.8,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7631578947368421,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.6911764705882353,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6266666666666667,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.9333333333333333,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9074074074074074,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8571428571428571,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.84375,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14092160761356354,
      "eval_overall_accuracy": 0.96556849145405,
      "eval_overall_f1": 0.8418156808803301,
      "eval_overall_precision": 0.8031496062992126,
      "eval_overall_recall": 0.884393063583815,
      "eval_runtime": 0.3246,
      "eval_samples_per_second": 576.135,
      "eval_steps_per_second": 9.243,
      "step": 6466
    },
    {
      "epoch": 62.0,
      "grad_norm": 0.4074559807777405,
      "learning_rate": 1.9e-05,
      "loss": 0.0475,
      "step": 6572
    },
    {
      "epoch": 62.0,
      "eval_LOCATION_f1": 0.7702702702702703,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7215189873417721,
      "eval_LOCATION_recall": 0.8260869565217391,
      "eval_ORGANIZATION_f1": 0.6923076923076924,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6521739130434783,
      "eval_ORGANIZATION_recall": 0.7377049180327869,
      "eval_PERSON_f1": 0.915360501567398,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8795180722891566,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8253968253968254,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8125,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.1439296454191208,
      "eval_overall_accuracy": 0.9633391132028735,
      "eval_overall_f1": 0.8269230769230771,
      "eval_overall_precision": 0.7879581151832461,
      "eval_overall_recall": 0.869942196531792,
      "eval_runtime": 0.3274,
      "eval_samples_per_second": 571.214,
      "eval_steps_per_second": 9.164,
      "step": 6572
    },
    {
      "epoch": 63.0,
      "grad_norm": 0.8321461081504822,
      "learning_rate": 1.85e-05,
      "loss": 0.0456,
      "step": 6678
    },
    {
      "epoch": 63.0,
      "eval_LOCATION_f1": 0.7866666666666667,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.7058823529411764,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.64,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9182389937106918,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8848484848484849,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.8059701492537314,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7714285714285715,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8571428571428571,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.84375,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15032930672168732,
      "eval_overall_accuracy": 0.9630914045082982,
      "eval_overall_f1": 0.8365122615803814,
      "eval_overall_precision": 0.7912371134020618,
      "eval_overall_recall": 0.8872832369942196,
      "eval_runtime": 0.325,
      "eval_samples_per_second": 575.43,
      "eval_steps_per_second": 9.231,
      "step": 6678
    },
    {
      "epoch": 64.0,
      "grad_norm": 1.6446304321289062,
      "learning_rate": 1.8e-05,
      "loss": 0.0469,
      "step": 6784
    },
    {
      "epoch": 64.0,
      "eval_LOCATION_f1": 0.7733333333333333,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7160493827160493,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.6805555555555556,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.5903614457831325,
      "eval_ORGANIZATION_recall": 0.8032786885245902,
      "eval_PERSON_f1": 0.915360501567398,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8795180722891566,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15162502229213715,
      "eval_overall_accuracy": 0.9618528610354223,
      "eval_overall_f1": 0.8230563002680965,
      "eval_overall_precision": 0.7675,
      "eval_overall_recall": 0.8872832369942196,
      "eval_runtime": 0.3263,
      "eval_samples_per_second": 573.107,
      "eval_steps_per_second": 9.194,
      "step": 6784
    },
    {
      "epoch": 65.0,
      "grad_norm": 0.5187699794769287,
      "learning_rate": 1.75e-05,
      "loss": 0.0461,
      "step": 6890
    },
    {
      "epoch": 65.0,
      "eval_LOCATION_f1": 0.7972972972972974,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7468354430379747,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6950354609929078,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6125,
      "eval_ORGANIZATION_recall": 0.8032786885245902,
      "eval_PERSON_f1": 0.915360501567398,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8795180722891566,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15085075795650482,
      "eval_overall_accuracy": 0.9623482784245727,
      "eval_overall_f1": 0.8313090418353577,
      "eval_overall_precision": 0.779746835443038,
      "eval_overall_recall": 0.8901734104046243,
      "eval_runtime": 0.3272,
      "eval_samples_per_second": 571.563,
      "eval_steps_per_second": 9.169,
      "step": 6890
    },
    {
      "epoch": 66.0,
      "grad_norm": 0.37507164478302,
      "learning_rate": 1.7000000000000003e-05,
      "loss": 0.0458,
      "step": 6996
    },
    {
      "epoch": 66.0,
      "eval_LOCATION_f1": 0.7919463087248322,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7375,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6666666666666667,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6081081081081081,
      "eval_ORGANIZATION_recall": 0.7377049180327869,
      "eval_PERSON_f1": 0.9182389937106918,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8848484848484849,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.84375,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8181818181818182,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14628949761390686,
      "eval_overall_accuracy": 0.9633391132028735,
      "eval_overall_f1": 0.8283378746594006,
      "eval_overall_precision": 0.7835051546391752,
      "eval_overall_recall": 0.8786127167630058,
      "eval_runtime": 0.3263,
      "eval_samples_per_second": 573.052,
      "eval_steps_per_second": 9.193,
      "step": 6996
    },
    {
      "epoch": 67.0,
      "grad_norm": 1.2049661874771118,
      "learning_rate": 1.65e-05,
      "loss": 0.046,
      "step": 7102
    },
    {
      "epoch": 67.0,
      "eval_LOCATION_f1": 0.8108108108108109,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6911764705882353,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6266666666666667,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.9245283018867925,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8909090909090909,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7826086956521738,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7297297297297297,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14728553593158722,
      "eval_overall_accuracy": 0.9635868218974486,
      "eval_overall_f1": 0.8369565217391304,
      "eval_overall_precision": 0.7897435897435897,
      "eval_overall_recall": 0.8901734104046243,
      "eval_runtime": 0.3262,
      "eval_samples_per_second": 573.352,
      "eval_steps_per_second": 9.198,
      "step": 7102
    },
    {
      "epoch": 68.0,
      "grad_norm": 0.525566577911377,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.0427,
      "step": 7208
    },
    {
      "epoch": 68.0,
      "eval_LOCATION_f1": 0.778523489932886,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.725,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.676470588235294,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6133333333333333,
      "eval_ORGANIZATION_recall": 0.7540983606557377,
      "eval_PERSON_f1": 0.915360501567398,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8795180722891566,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14847737550735474,
      "eval_overall_accuracy": 0.9643299479811741,
      "eval_overall_f1": 0.8238482384823848,
      "eval_overall_precision": 0.7755102040816326,
      "eval_overall_recall": 0.8786127167630058,
      "eval_runtime": 0.3256,
      "eval_samples_per_second": 574.396,
      "eval_steps_per_second": 9.215,
      "step": 7208
    },
    {
      "epoch": 69.0,
      "grad_norm": 2.024226665496826,
      "learning_rate": 1.55e-05,
      "loss": 0.0436,
      "step": 7314
    },
    {
      "epoch": 69.0,
      "eval_LOCATION_f1": 0.8243243243243245,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7721518987341772,
      "eval_LOCATION_recall": 0.8840579710144928,
      "eval_ORGANIZATION_f1": 0.6962962962962963,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6351351351351351,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.9333333333333333,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9074074074074074,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.84375,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8181818181818182,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14697425067424774,
      "eval_overall_accuracy": 0.9650730740648997,
      "eval_overall_f1": 0.8465753424657535,
      "eval_overall_precision": 0.8046875,
      "eval_overall_recall": 0.8930635838150289,
      "eval_runtime": 0.3268,
      "eval_samples_per_second": 572.188,
      "eval_steps_per_second": 9.179,
      "step": 7314
    },
    {
      "epoch": 70.0,
      "grad_norm": 2.120527982711792,
      "learning_rate": 1.5e-05,
      "loss": 0.0443,
      "step": 7420
    },
    {
      "epoch": 70.0,
      "eval_LOCATION_f1": 0.7919463087248322,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7375,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.711111111111111,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6486486486486487,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9274447949526814,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8963414634146342,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.8059701492537314,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7714285714285715,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15468451380729675,
      "eval_overall_accuracy": 0.964082239286599,
      "eval_overall_f1": 0.8403819918144612,
      "eval_overall_precision": 0.7958656330749354,
      "eval_overall_recall": 0.8901734104046243,
      "eval_runtime": 0.3267,
      "eval_samples_per_second": 572.368,
      "eval_steps_per_second": 9.182,
      "step": 7420
    },
    {
      "epoch": 71.0,
      "grad_norm": 0.9601520895957947,
      "learning_rate": 1.45e-05,
      "loss": 0.0422,
      "step": 7526
    },
    {
      "epoch": 71.0,
      "eval_LOCATION_f1": 0.7733333333333333,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7160493827160493,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.7007299270072993,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.631578947368421,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9240506329113924,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8957055214723927,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.8059701492537314,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7714285714285715,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.1504492461681366,
      "eval_overall_accuracy": 0.9648253653703245,
      "eval_overall_f1": 0.8315217391304348,
      "eval_overall_precision": 0.7846153846153846,
      "eval_overall_recall": 0.884393063583815,
      "eval_runtime": 0.3282,
      "eval_samples_per_second": 569.76,
      "eval_steps_per_second": 9.141,
      "step": 7526
    },
    {
      "epoch": 72.0,
      "grad_norm": 0.6983500123023987,
      "learning_rate": 1.4000000000000001e-05,
      "loss": 0.042,
      "step": 7632
    },
    {
      "epoch": 72.0,
      "eval_LOCATION_f1": 0.7891156462585034,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7435897435897436,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.711111111111111,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6486486486486487,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9333333333333333,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9074074074074074,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.8059701492537314,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7714285714285715,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15169183909893036,
      "eval_overall_accuracy": 0.9650730740648997,
      "eval_overall_f1": 0.8410958904109589,
      "eval_overall_precision": 0.7994791666666666,
      "eval_overall_recall": 0.8872832369942196,
      "eval_runtime": 0.3258,
      "eval_samples_per_second": 573.921,
      "eval_steps_per_second": 9.207,
      "step": 7632
    },
    {
      "epoch": 73.0,
      "grad_norm": 1.0813753604888916,
      "learning_rate": 1.3500000000000001e-05,
      "loss": 0.0409,
      "step": 7738
    },
    {
      "epoch": 73.0,
      "eval_LOCATION_f1": 0.7891156462585034,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7435897435897436,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.6716417910447762,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6164383561643836,
      "eval_ORGANIZATION_recall": 0.7377049180327869,
      "eval_PERSON_f1": 0.9333333333333333,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9074074074074074,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.8059701492537314,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7714285714285715,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15077091753482819,
      "eval_overall_accuracy": 0.9638345305920237,
      "eval_overall_f1": 0.8340192043895747,
      "eval_overall_precision": 0.793733681462141,
      "eval_overall_recall": 0.8786127167630058,
      "eval_runtime": 0.3273,
      "eval_samples_per_second": 571.428,
      "eval_steps_per_second": 9.167,
      "step": 7738
    },
    {
      "epoch": 74.0,
      "grad_norm": 0.23750807344913483,
      "learning_rate": 1.3000000000000001e-05,
      "loss": 0.0423,
      "step": 7844
    },
    {
      "epoch": 74.0,
      "eval_LOCATION_f1": 0.7919463087248322,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7375,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6944444444444445,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6024096385542169,
      "eval_ORGANIZATION_recall": 0.819672131147541,
      "eval_PERSON_f1": 0.9303797468354431,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.901840490797546,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.8059701492537314,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7714285714285715,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8059701492537312,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.75,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15902836620807648,
      "eval_overall_accuracy": 0.9616051523408472,
      "eval_overall_f1": 0.8344549125168238,
      "eval_overall_precision": 0.7808564231738035,
      "eval_overall_recall": 0.8959537572254336,
      "eval_runtime": 0.3269,
      "eval_samples_per_second": 571.967,
      "eval_steps_per_second": 9.176,
      "step": 7844
    },
    {
      "epoch": 75.0,
      "grad_norm": 1.3622193336486816,
      "learning_rate": 1.25e-05,
      "loss": 0.0425,
      "step": 7950
    },
    {
      "epoch": 75.0,
      "eval_LOCATION_f1": 0.7866666666666667,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.6715328467153285,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6052631578947368,
      "eval_ORGANIZATION_recall": 0.7540983606557377,
      "eval_PERSON_f1": 0.9274447949526814,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8963414634146342,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.8059701492537314,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7714285714285715,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.84375,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8181818181818182,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14874456822872162,
      "eval_overall_accuracy": 0.964082239286599,
      "eval_overall_f1": 0.8326530612244898,
      "eval_overall_precision": 0.7866323907455013,
      "eval_overall_recall": 0.884393063583815,
      "eval_runtime": 0.3268,
      "eval_samples_per_second": 572.253,
      "eval_steps_per_second": 9.181,
      "step": 7950
    },
    {
      "epoch": 76.0,
      "grad_norm": 1.7681535482406616,
      "learning_rate": 1.2e-05,
      "loss": 0.0398,
      "step": 8056
    },
    {
      "epoch": 76.0,
      "eval_LOCATION_f1": 0.7919463087248322,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7375,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.7194244604316549,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6410256410256411,
      "eval_ORGANIZATION_recall": 0.819672131147541,
      "eval_PERSON_f1": 0.9303797468354431,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.901840490797546,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7536231884057971,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7027027027027027,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15581201016902924,
      "eval_overall_accuracy": 0.9633391132028735,
      "eval_overall_f1": 0.8373983739837398,
      "eval_overall_precision": 0.7882653061224489,
      "eval_overall_recall": 0.8930635838150289,
      "eval_runtime": 0.3274,
      "eval_samples_per_second": 571.16,
      "eval_steps_per_second": 9.163,
      "step": 8056
    },
    {
      "epoch": 77.0,
      "grad_norm": 0.41535985469818115,
      "learning_rate": 1.1500000000000002e-05,
      "loss": 0.0426,
      "step": 8162
    },
    {
      "epoch": 77.0,
      "eval_LOCATION_f1": 0.8108108108108109,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.7092198581560285,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.625,
      "eval_ORGANIZATION_recall": 0.819672131147541,
      "eval_PERSON_f1": 0.9274447949526814,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8963414634146342,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8059701492537312,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.75,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15663787722587585,
      "eval_overall_accuracy": 0.9628436958137231,
      "eval_overall_f1": 0.8394062078272604,
      "eval_overall_precision": 0.7873417721518987,
      "eval_overall_recall": 0.8988439306358381,
      "eval_runtime": 0.3256,
      "eval_samples_per_second": 574.372,
      "eval_steps_per_second": 9.215,
      "step": 8162
    },
    {
      "epoch": 78.0,
      "grad_norm": 2.316864490509033,
      "learning_rate": 1.1000000000000001e-05,
      "loss": 0.0416,
      "step": 8268
    },
    {
      "epoch": 78.0,
      "eval_LOCATION_f1": 0.8243243243243245,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7721518987341772,
      "eval_LOCATION_recall": 0.8840579710144928,
      "eval_ORGANIZATION_f1": 0.7007299270072993,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.631578947368421,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9333333333333333,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9074074074074074,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7647058823529411,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7222222222222222,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15233848989009857,
      "eval_overall_accuracy": 0.9628436958137231,
      "eval_overall_f1": 0.8419618528610355,
      "eval_overall_precision": 0.7963917525773195,
      "eval_overall_recall": 0.8930635838150289,
      "eval_runtime": 0.325,
      "eval_samples_per_second": 575.423,
      "eval_steps_per_second": 9.231,
      "step": 8268
    },
    {
      "epoch": 79.0,
      "grad_norm": 0.4779236614704132,
      "learning_rate": 1.05e-05,
      "loss": 0.0415,
      "step": 8374
    },
    {
      "epoch": 79.0,
      "eval_LOCATION_f1": 0.8137931034482757,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7763157894736842,
      "eval_LOCATION_recall": 0.855072463768116,
      "eval_ORGANIZATION_f1": 0.7164179104477612,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6575342465753424,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9333333333333333,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9074074074074074,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8125,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7878787878787878,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.1466650515794754,
      "eval_overall_accuracy": 0.9638345305920237,
      "eval_overall_f1": 0.8457300275482094,
      "eval_overall_precision": 0.8078947368421052,
      "eval_overall_recall": 0.8872832369942196,
      "eval_runtime": 0.3264,
      "eval_samples_per_second": 572.831,
      "eval_steps_per_second": 9.19,
      "step": 8374
    },
    {
      "epoch": 80.0,
      "grad_norm": 0.4587147533893585,
      "learning_rate": 1e-05,
      "loss": 0.0389,
      "step": 8480
    },
    {
      "epoch": 80.0,
      "eval_LOCATION_f1": 0.8108108108108109,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6956521739130435,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6233766233766234,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9333333333333333,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9074074074074074,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7536231884057971,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7027027027027027,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15246228873729706,
      "eval_overall_accuracy": 0.9628436958137231,
      "eval_overall_f1": 0.8369565217391304,
      "eval_overall_precision": 0.7897435897435897,
      "eval_overall_recall": 0.8901734104046243,
      "eval_runtime": 0.325,
      "eval_samples_per_second": 575.441,
      "eval_steps_per_second": 9.232,
      "step": 8480
    },
    {
      "epoch": 81.0,
      "grad_norm": 0.5301854014396667,
      "learning_rate": 9.5e-06,
      "loss": 0.0378,
      "step": 8586
    },
    {
      "epoch": 81.0,
      "eval_LOCATION_f1": 0.8108108108108109,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.7058823529411764,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.64,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9303797468354431,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.901840490797546,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7536231884057971,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7027027027027027,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.7999999999999999,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7647058823529411,
      "eval_TIME_recall": 0.8387096774193549,
      "eval_loss": 0.1581680327653885,
      "eval_overall_accuracy": 0.9625959871191478,
      "eval_overall_f1": 0.8365122615803814,
      "eval_overall_precision": 0.7912371134020618,
      "eval_overall_recall": 0.8872832369942196,
      "eval_runtime": 0.3261,
      "eval_samples_per_second": 573.45,
      "eval_steps_per_second": 9.2,
      "step": 8586
    },
    {
      "epoch": 82.0,
      "grad_norm": 0.4550780951976776,
      "learning_rate": 9e-06,
      "loss": 0.041,
      "step": 8692
    },
    {
      "epoch": 82.0,
      "eval_LOCATION_f1": 0.8108108108108109,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.7058823529411764,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.64,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9367088607594937,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9079754601226994,
      "eval_PERSON_recall": 0.9673202614379085,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.84375,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8181818181818182,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.1535988450050354,
      "eval_overall_accuracy": 0.9643299479811741,
      "eval_overall_f1": 0.8469945355191256,
      "eval_overall_precision": 0.8031088082901554,
      "eval_overall_recall": 0.8959537572254336,
      "eval_runtime": 0.3257,
      "eval_samples_per_second": 574.174,
      "eval_steps_per_second": 9.211,
      "step": 8692
    },
    {
      "epoch": 83.0,
      "grad_norm": 0.37215691804885864,
      "learning_rate": 8.500000000000002e-06,
      "loss": 0.0385,
      "step": 8798
    },
    {
      "epoch": 83.0,
      "eval_LOCATION_f1": 0.821917808219178,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7792207792207793,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6906474820143885,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6153846153846154,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9211356466876972,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8902439024390244,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.84375,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8181818181818182,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.1516975313425064,
      "eval_overall_accuracy": 0.9635868218974486,
      "eval_overall_f1": 0.8392370572207084,
      "eval_overall_precision": 0.7938144329896907,
      "eval_overall_recall": 0.8901734104046243,
      "eval_runtime": 0.3257,
      "eval_samples_per_second": 574.178,
      "eval_steps_per_second": 9.211,
      "step": 8798
    },
    {
      "epoch": 84.0,
      "grad_norm": 1.83058762550354,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.0386,
      "step": 8904
    },
    {
      "epoch": 84.0,
      "eval_LOCATION_f1": 0.8163265306122449,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7692307692307693,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.7101449275362318,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6363636363636364,
      "eval_ORGANIZATION_recall": 0.8032786885245902,
      "eval_PERSON_f1": 0.9303797468354431,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.901840490797546,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8571428571428571,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.84375,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15184840559959412,
      "eval_overall_accuracy": 0.96556849145405,
      "eval_overall_f1": 0.8469945355191256,
      "eval_overall_precision": 0.8031088082901554,
      "eval_overall_recall": 0.8959537572254336,
      "eval_runtime": 0.3257,
      "eval_samples_per_second": 574.084,
      "eval_steps_per_second": 9.21,
      "step": 8904
    },
    {
      "epoch": 85.0,
      "grad_norm": 2.1711184978485107,
      "learning_rate": 7.5e-06,
      "loss": 0.0397,
      "step": 9010
    },
    {
      "epoch": 85.0,
      "eval_LOCATION_f1": 0.8243243243243245,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7721518987341772,
      "eval_LOCATION_recall": 0.8840579710144928,
      "eval_ORGANIZATION_f1": 0.6911764705882353,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6266666666666667,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.9337539432176657,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9024390243902439,
      "eval_PERSON_recall": 0.9673202614379085,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.84375,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8181818181818182,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15403476357460022,
      "eval_overall_accuracy": 0.9633391132028735,
      "eval_overall_f1": 0.8458390177353343,
      "eval_overall_precision": 0.8010335917312662,
      "eval_overall_recall": 0.8959537572254336,
      "eval_runtime": 0.3269,
      "eval_samples_per_second": 572.126,
      "eval_steps_per_second": 9.178,
      "step": 9010
    },
    {
      "epoch": 86.0,
      "grad_norm": 1.9255669116973877,
      "learning_rate": 7.000000000000001e-06,
      "loss": 0.038,
      "step": 9116
    },
    {
      "epoch": 86.0,
      "eval_LOCATION_f1": 0.8108108108108109,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6956521739130435,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6233766233766234,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9367088607594937,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9079754601226994,
      "eval_PERSON_recall": 0.9673202614379085,
      "eval_QUANTITY_f1": 0.7826086956521738,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7297297297297297,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15543720126152039,
      "eval_overall_accuracy": 0.9638345305920237,
      "eval_overall_f1": 0.8412483039348712,
      "eval_overall_precision": 0.7928388746803069,
      "eval_overall_recall": 0.8959537572254336,
      "eval_runtime": 0.326,
      "eval_samples_per_second": 573.621,
      "eval_steps_per_second": 9.202,
      "step": 9116
    },
    {
      "epoch": 87.0,
      "grad_norm": 0.8977526426315308,
      "learning_rate": 6.5000000000000004e-06,
      "loss": 0.0379,
      "step": 9222
    },
    {
      "epoch": 87.0,
      "eval_LOCATION_f1": 0.8243243243243245,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7721518987341772,
      "eval_LOCATION_recall": 0.8840579710144928,
      "eval_ORGANIZATION_f1": 0.6906474820143885,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6153846153846154,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9367088607594937,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9079754601226994,
      "eval_PERSON_recall": 0.9673202614379085,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.1552109271287918,
      "eval_overall_accuracy": 0.9635868218974486,
      "eval_overall_f1": 0.8451086956521738,
      "eval_overall_precision": 0.7974358974358975,
      "eval_overall_recall": 0.8988439306358381,
      "eval_runtime": 0.3263,
      "eval_samples_per_second": 573.095,
      "eval_steps_per_second": 9.194,
      "step": 9222
    },
    {
      "epoch": 88.0,
      "grad_norm": 0.9438088536262512,
      "learning_rate": 6e-06,
      "loss": 0.0363,
      "step": 9328
    },
    {
      "epoch": 88.0,
      "eval_LOCATION_f1": 0.8108108108108109,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.712121212121212,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6619718309859155,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.9303797468354431,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.901840490797546,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15475918352603912,
      "eval_overall_accuracy": 0.9638345305920237,
      "eval_overall_f1": 0.8449931412894376,
      "eval_overall_precision": 0.804177545691906,
      "eval_overall_recall": 0.8901734104046243,
      "eval_runtime": 0.3257,
      "eval_samples_per_second": 574.195,
      "eval_steps_per_second": 9.212,
      "step": 9328
    },
    {
      "epoch": 89.0,
      "grad_norm": 0.8352949023246765,
      "learning_rate": 5.500000000000001e-06,
      "loss": 0.0376,
      "step": 9434
    },
    {
      "epoch": 89.0,
      "eval_LOCATION_f1": 0.8163265306122449,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7692307692307693,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6917293233082706,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6388888888888888,
      "eval_ORGANIZATION_recall": 0.7540983606557377,
      "eval_PERSON_f1": 0.9269841269841269,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9012345679012346,
      "eval_PERSON_recall": 0.954248366013072,
      "eval_QUANTITY_f1": 0.8059701492537314,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7714285714285715,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.84375,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.8181818181818182,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.14811310172080994,
      "eval_overall_accuracy": 0.96556849145405,
      "eval_overall_f1": 0.8429752066115701,
      "eval_overall_precision": 0.8052631578947368,
      "eval_overall_recall": 0.884393063583815,
      "eval_runtime": 0.326,
      "eval_samples_per_second": 573.674,
      "eval_steps_per_second": 9.203,
      "step": 9434
    },
    {
      "epoch": 90.0,
      "grad_norm": 1.0256093740463257,
      "learning_rate": 5e-06,
      "loss": 0.0348,
      "step": 9540
    },
    {
      "epoch": 90.0,
      "eval_LOCATION_f1": 0.8108108108108109,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6956521739130435,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6233766233766234,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9333333333333333,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9074074074074074,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.1592727154493332,
      "eval_overall_accuracy": 0.9635868218974486,
      "eval_overall_f1": 0.8419618528610355,
      "eval_overall_precision": 0.7963917525773195,
      "eval_overall_recall": 0.8930635838150289,
      "eval_runtime": 0.3258,
      "eval_samples_per_second": 573.968,
      "eval_steps_per_second": 9.208,
      "step": 9540
    },
    {
      "epoch": 91.0,
      "grad_norm": 0.45585766434669495,
      "learning_rate": 4.5e-06,
      "loss": 0.0362,
      "step": 9646
    },
    {
      "epoch": 91.0,
      "eval_LOCATION_f1": 0.7945205479452054,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7532467532467533,
      "eval_LOCATION_recall": 0.8405797101449275,
      "eval_ORGANIZATION_f1": 0.7014925373134328,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6438356164383562,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.9274447949526814,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8963414634146342,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.8059701492537314,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7714285714285715,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15180537104606628,
      "eval_overall_accuracy": 0.9645776566757494,
      "eval_overall_f1": 0.839506172839506,
      "eval_overall_precision": 0.7989556135770235,
      "eval_overall_recall": 0.884393063583815,
      "eval_runtime": 0.3254,
      "eval_samples_per_second": 574.629,
      "eval_steps_per_second": 9.219,
      "step": 9646
    },
    {
      "epoch": 92.0,
      "grad_norm": 0.49642303586006165,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0383,
      "step": 9752
    },
    {
      "epoch": 92.0,
      "eval_LOCATION_f1": 0.8243243243243245,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7721518987341772,
      "eval_LOCATION_recall": 0.8840579710144928,
      "eval_ORGANIZATION_f1": 0.6906474820143885,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6153846153846154,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9274447949526814,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.8963414634146342,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15918388962745667,
      "eval_overall_accuracy": 0.9623482784245727,
      "eval_overall_f1": 0.840108401084011,
      "eval_overall_precision": 0.7908163265306123,
      "eval_overall_recall": 0.8959537572254336,
      "eval_runtime": 0.3255,
      "eval_samples_per_second": 574.5,
      "eval_steps_per_second": 9.217,
      "step": 9752
    },
    {
      "epoch": 93.0,
      "grad_norm": 1.612742304801941,
      "learning_rate": 3.5000000000000004e-06,
      "loss": 0.0363,
      "step": 9858
    },
    {
      "epoch": 93.0,
      "eval_LOCATION_f1": 0.8187919463087248,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.7625,
      "eval_LOCATION_recall": 0.8840579710144928,
      "eval_ORGANIZATION_f1": 0.6901408450704226,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6049382716049383,
      "eval_ORGANIZATION_recall": 0.8032786885245902,
      "eval_PERSON_f1": 0.9333333333333333,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9074074074074074,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7536231884057971,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7027027027027027,
      "eval_QUANTITY_recall": 0.8125,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.1625114232301712,
      "eval_overall_accuracy": 0.961357443646272,
      "eval_overall_f1": 0.8367071524966261,
      "eval_overall_precision": 0.7848101265822784,
      "eval_overall_recall": 0.8959537572254336,
      "eval_runtime": 0.3253,
      "eval_samples_per_second": 574.916,
      "eval_steps_per_second": 9.223,
      "step": 9858
    },
    {
      "epoch": 94.0,
      "grad_norm": 0.48744139075279236,
      "learning_rate": 3e-06,
      "loss": 0.0377,
      "step": 9964
    },
    {
      "epoch": 94.0,
      "eval_LOCATION_f1": 0.8053691275167786,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.75,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.7014925373134328,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6438356164383562,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.9333333333333333,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9074074074074074,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.8059701492537314,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.7714285714285715,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15427272021770477,
      "eval_overall_accuracy": 0.9645776566757494,
      "eval_overall_f1": 0.8426812585499317,
      "eval_overall_precision": 0.8,
      "eval_overall_recall": 0.8901734104046243,
      "eval_runtime": 0.3278,
      "eval_samples_per_second": 570.486,
      "eval_steps_per_second": 9.152,
      "step": 9964
    },
    {
      "epoch": 95.0,
      "grad_norm": 1.4551924467086792,
      "learning_rate": 2.5e-06,
      "loss": 0.0362,
      "step": 10070
    },
    {
      "epoch": 95.0,
      "eval_LOCATION_f1": 0.8108108108108109,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6956521739130435,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6233766233766234,
      "eval_ORGANIZATION_recall": 0.7868852459016393,
      "eval_PERSON_f1": 0.9333333333333333,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9074074074074074,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.1584785431623459,
      "eval_overall_accuracy": 0.9630914045082982,
      "eval_overall_f1": 0.8408163265306122,
      "eval_overall_precision": 0.794344473007712,
      "eval_overall_recall": 0.8930635838150289,
      "eval_runtime": 0.3268,
      "eval_samples_per_second": 572.147,
      "eval_steps_per_second": 9.179,
      "step": 10070
    },
    {
      "epoch": 96.0,
      "grad_norm": 0.7348619103431702,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0351,
      "step": 10176
    },
    {
      "epoch": 96.0,
      "eval_LOCATION_f1": 0.8299319727891156,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.782051282051282,
      "eval_LOCATION_recall": 0.8840579710144928,
      "eval_ORGANIZATION_f1": 0.6911764705882353,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6266666666666667,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.9333333333333333,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9074074074074074,
      "eval_PERSON_recall": 0.9607843137254902,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15524066984653473,
      "eval_overall_accuracy": 0.964082239286599,
      "eval_overall_f1": 0.8442622950819672,
      "eval_overall_precision": 0.8005181347150259,
      "eval_overall_recall": 0.8930635838150289,
      "eval_runtime": 0.3255,
      "eval_samples_per_second": 574.586,
      "eval_steps_per_second": 9.218,
      "step": 10176
    },
    {
      "epoch": 97.0,
      "grad_norm": 0.6167792677879333,
      "learning_rate": 1.5e-06,
      "loss": 0.036,
      "step": 10282
    },
    {
      "epoch": 97.0,
      "eval_LOCATION_f1": 0.8299319727891156,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.782051282051282,
      "eval_LOCATION_recall": 0.8840579710144928,
      "eval_ORGANIZATION_f1": 0.6962962962962963,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6351351351351351,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.9367088607594937,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9079754601226994,
      "eval_PERSON_recall": 0.9673202614379085,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8307692307692308,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7941176470588235,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.152723491191864,
      "eval_overall_accuracy": 0.9650730740648997,
      "eval_overall_f1": 0.8481532147742818,
      "eval_overall_precision": 0.8051948051948052,
      "eval_overall_recall": 0.8959537572254336,
      "eval_runtime": 0.3255,
      "eval_samples_per_second": 574.541,
      "eval_steps_per_second": 9.217,
      "step": 10282
    },
    {
      "epoch": 98.0,
      "grad_norm": 0.26082101464271545,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0368,
      "step": 10388
    },
    {
      "epoch": 98.0,
      "eval_LOCATION_f1": 0.8108108108108109,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6962962962962963,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6351351351351351,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.9367088607594937,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9079754601226994,
      "eval_PERSON_recall": 0.9673202614379085,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15440693497657776,
      "eval_overall_accuracy": 0.9638345305920237,
      "eval_overall_f1": 0.8431105047748976,
      "eval_overall_precision": 0.7984496124031008,
      "eval_overall_recall": 0.8930635838150289,
      "eval_runtime": 0.3264,
      "eval_samples_per_second": 572.99,
      "eval_steps_per_second": 9.192,
      "step": 10388
    },
    {
      "epoch": 99.0,
      "grad_norm": 1.4792131185531616,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0362,
      "step": 10494
    },
    {
      "epoch": 99.0,
      "eval_LOCATION_f1": 0.8108108108108109,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6962962962962963,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6351351351351351,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.9367088607594937,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9079754601226994,
      "eval_PERSON_recall": 0.9673202614379085,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.15458133816719055,
      "eval_overall_accuracy": 0.9635868218974486,
      "eval_overall_f1": 0.8431105047748976,
      "eval_overall_precision": 0.7984496124031008,
      "eval_overall_recall": 0.8930635838150289,
      "eval_runtime": 0.3316,
      "eval_samples_per_second": 563.856,
      "eval_steps_per_second": 9.046,
      "step": 10494
    },
    {
      "epoch": 100.0,
      "grad_norm": 3.9763753414154053,
      "learning_rate": 0.0,
      "loss": 0.034,
      "step": 10600
    },
    {
      "epoch": 100.0,
      "eval_LOCATION_f1": 0.8108108108108109,
      "eval_LOCATION_number": 69,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8695652173913043,
      "eval_ORGANIZATION_f1": 0.6962962962962963,
      "eval_ORGANIZATION_number": 61,
      "eval_ORGANIZATION_precision": 0.6351351351351351,
      "eval_ORGANIZATION_recall": 0.7704918032786885,
      "eval_PERSON_f1": 0.9367088607594937,
      "eval_PERSON_number": 153,
      "eval_PERSON_precision": 0.9079754601226994,
      "eval_PERSON_recall": 0.9673202614379085,
      "eval_QUANTITY_f1": 0.7941176470588235,
      "eval_QUANTITY_number": 32,
      "eval_QUANTITY_precision": 0.75,
      "eval_QUANTITY_recall": 0.84375,
      "eval_TIME_f1": 0.8181818181818182,
      "eval_TIME_number": 31,
      "eval_TIME_precision": 0.7714285714285715,
      "eval_TIME_recall": 0.8709677419354839,
      "eval_loss": 0.1546700894832611,
      "eval_overall_accuracy": 0.9635868218974486,
      "eval_overall_f1": 0.8431105047748976,
      "eval_overall_precision": 0.7984496124031008,
      "eval_overall_recall": 0.8930635838150289,
      "eval_runtime": 0.3265,
      "eval_samples_per_second": 572.826,
      "eval_steps_per_second": 9.19,
      "step": 10600
    },
    {
      "epoch": 100.0,
      "step": 10600,
      "total_flos": 5088556202474382.0,
      "train_loss": 0.0811429582002028,
      "train_runtime": 569.0335,
      "train_samples_per_second": 296.468,
      "train_steps_per_second": 18.628
    }
  ],
  "logging_steps": 500,
  "max_steps": 10600,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 100,
  "save_steps": 500,
  "total_flos": 5088556202474382.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}