trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 100.0,
  "eval_steps": 500,
  "global_step": 9600,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 1.4548790454864502,
      "learning_rate": 4.9500000000000004e-05,
      "loss": 1.1704,
      "step": 96
    },
    {
      "epoch": 1.0,
      "eval_LOCATION_f1": 0.0,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.0,
      "eval_LOCATION_recall": 0.0,
      "eval_ORGANIZATION_f1": 0.0,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.0,
      "eval_ORGANIZATION_recall": 0.0,
      "eval_PERSON_f1": 0.0,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.0,
      "eval_PERSON_recall": 0.0,
      "eval_loss": 0.708549976348877,
      "eval_overall_accuracy": 0.8342935528120713,
      "eval_overall_f1": 0.0,
      "eval_overall_precision": 0.0,
      "eval_overall_recall": 0.0,
      "eval_runtime": 0.2796,
      "eval_samples_per_second": 608.047,
      "eval_steps_per_second": 10.73,
      "step": 96
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.531407117843628,
      "learning_rate": 4.9e-05,
      "loss": 0.668,
      "step": 192
    },
    {
      "epoch": 2.0,
      "eval_LOCATION_f1": 0.0,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.0,
      "eval_LOCATION_recall": 0.0,
      "eval_ORGANIZATION_f1": 0.011904761904761906,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.5,
      "eval_ORGANIZATION_recall": 0.006024096385542169,
      "eval_PERSON_f1": 0.0,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.0,
      "eval_PERSON_recall": 0.0,
      "eval_loss": 0.5722576975822449,
      "eval_overall_accuracy": 0.8348422496570644,
      "eval_overall_f1": 0.00495049504950495,
      "eval_overall_precision": 0.3333333333333333,
      "eval_overall_recall": 0.0024937655860349127,
      "eval_runtime": 0.2814,
      "eval_samples_per_second": 604.079,
      "eval_steps_per_second": 10.66,
      "step": 192
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.9092353582382202,
      "learning_rate": 4.85e-05,
      "loss": 0.5537,
      "step": 288
    },
    {
      "epoch": 3.0,
      "eval_LOCATION_f1": 0.0,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.0,
      "eval_LOCATION_recall": 0.0,
      "eval_ORGANIZATION_f1": 0.10526315789473684,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.4166666666666667,
      "eval_ORGANIZATION_recall": 0.060240963855421686,
      "eval_PERSON_f1": 0.09090909090909091,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.23529411764705882,
      "eval_PERSON_recall": 0.056338028169014086,
      "eval_loss": 0.44938454031944275,
      "eval_overall_accuracy": 0.8455418381344307,
      "eval_overall_f1": 0.07809110629067245,
      "eval_overall_precision": 0.3,
      "eval_overall_recall": 0.04488778054862843,
      "eval_runtime": 0.2836,
      "eval_samples_per_second": 599.368,
      "eval_steps_per_second": 10.577,
      "step": 288
    },
    {
      "epoch": 4.0,
      "grad_norm": 1.0577952861785889,
      "learning_rate": 4.8e-05,
      "loss": 0.4382,
      "step": 384
    },
    {
      "epoch": 4.0,
      "eval_LOCATION_f1": 0.10434782608695652,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.2727272727272727,
      "eval_LOCATION_recall": 0.06451612903225806,
      "eval_ORGANIZATION_f1": 0.31724137931034485,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.3709677419354839,
      "eval_ORGANIZATION_recall": 0.27710843373493976,
      "eval_PERSON_f1": 0.4230769230769231,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.38823529411764707,
      "eval_PERSON_recall": 0.4647887323943662,
      "eval_loss": 0.3280525207519531,
      "eval_overall_accuracy": 0.8883401920438957,
      "eval_overall_f1": 0.32914923291492326,
      "eval_overall_precision": 0.37341772151898733,
      "eval_overall_recall": 0.2942643391521197,
      "eval_runtime": 0.2879,
      "eval_samples_per_second": 590.4,
      "eval_steps_per_second": 10.419,
      "step": 384
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.9427235126495361,
      "learning_rate": 4.75e-05,
      "loss": 0.32,
      "step": 480
    },
    {
      "epoch": 5.0,
      "eval_LOCATION_f1": 0.3312883435582822,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.38571428571428573,
      "eval_LOCATION_recall": 0.2903225806451613,
      "eval_ORGANIZATION_f1": 0.5650969529085873,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.5230769230769231,
      "eval_ORGANIZATION_recall": 0.6144578313253012,
      "eval_PERSON_f1": 0.6498422712933754,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.5885714285714285,
      "eval_PERSON_recall": 0.7253521126760564,
      "eval_loss": 0.2350081503391266,
      "eval_overall_accuracy": 0.9292181069958848,
      "eval_overall_f1": 0.5517241379310345,
      "eval_overall_precision": 0.5272727272727272,
      "eval_overall_recall": 0.5785536159600998,
      "eval_runtime": 0.2889,
      "eval_samples_per_second": 588.423,
      "eval_steps_per_second": 10.384,
      "step": 480
    },
    {
      "epoch": 6.0,
      "grad_norm": 1.7494196891784668,
      "learning_rate": 4.7e-05,
      "loss": 0.2426,
      "step": 576
    },
    {
      "epoch": 6.0,
      "eval_LOCATION_f1": 0.5775401069518716,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.574468085106383,
      "eval_LOCATION_recall": 0.5806451612903226,
      "eval_ORGANIZATION_f1": 0.6775067750677507,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.6157635467980296,
      "eval_ORGANIZATION_recall": 0.7530120481927711,
      "eval_PERSON_f1": 0.8208469055374593,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.7636363636363637,
      "eval_PERSON_recall": 0.8873239436619719,
      "eval_loss": 0.18390628695487976,
      "eval_overall_accuracy": 0.9511659807956104,
      "eval_overall_f1": 0.7068366164542296,
      "eval_overall_precision": 0.6601731601731602,
      "eval_overall_recall": 0.7605985037406484,
      "eval_runtime": 0.2856,
      "eval_samples_per_second": 595.176,
      "eval_steps_per_second": 10.503,
      "step": 576
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.9670518040657043,
      "learning_rate": 4.6500000000000005e-05,
      "loss": 0.1962,
      "step": 672
    },
    {
      "epoch": 7.0,
      "eval_LOCATION_f1": 0.7301587301587302,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.71875,
      "eval_LOCATION_recall": 0.7419354838709677,
      "eval_ORGANIZATION_f1": 0.7333333333333334,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.6804123711340206,
      "eval_ORGANIZATION_recall": 0.7951807228915663,
      "eval_PERSON_f1": 0.9292929292929293,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.8903225806451613,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.1462562382221222,
      "eval_overall_accuracy": 0.9618655692729767,
      "eval_overall_f1": 0.8014184397163121,
      "eval_overall_precision": 0.7617977528089888,
      "eval_overall_recall": 0.8453865336658354,
      "eval_runtime": 0.2846,
      "eval_samples_per_second": 597.237,
      "eval_steps_per_second": 10.539,
      "step": 672
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.7352226376533508,
      "learning_rate": 4.600000000000001e-05,
      "loss": 0.1696,
      "step": 768
    },
    {
      "epoch": 8.0,
      "eval_LOCATION_f1": 0.7894736842105262,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.7731958762886598,
      "eval_LOCATION_recall": 0.8064516129032258,
      "eval_ORGANIZATION_f1": 0.7727272727272727,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.7311827956989247,
      "eval_ORGANIZATION_recall": 0.8192771084337349,
      "eval_PERSON_f1": 0.9383561643835617,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9133333333333333,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.11997510492801666,
      "eval_overall_accuracy": 0.9681755829903979,
      "eval_overall_f1": 0.8345323741007196,
      "eval_overall_precision": 0.8036951501154734,
      "eval_overall_recall": 0.8678304239401496,
      "eval_runtime": 0.2879,
      "eval_samples_per_second": 590.554,
      "eval_steps_per_second": 10.422,
      "step": 768
    },
    {
      "epoch": 9.0,
      "grad_norm": 0.8205360174179077,
      "learning_rate": 4.55e-05,
      "loss": 0.1508,
      "step": 864
    },
    {
      "epoch": 9.0,
      "eval_LOCATION_f1": 0.8290155440414508,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8,
      "eval_LOCATION_recall": 0.8602150537634409,
      "eval_ORGANIZATION_f1": 0.7897727272727273,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.7473118279569892,
      "eval_ORGANIZATION_recall": 0.8373493975903614,
      "eval_PERSON_f1": 0.9387755102040817,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9078947368421053,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.10688301175832748,
      "eval_overall_accuracy": 0.9695473251028807,
      "eval_overall_f1": 0.8510131108462454,
      "eval_overall_precision": 0.815068493150685,
      "eval_overall_recall": 0.8902743142144638,
      "eval_runtime": 0.2939,
      "eval_samples_per_second": 578.353,
      "eval_steps_per_second": 10.206,
      "step": 864
    },
    {
      "epoch": 10.0,
      "grad_norm": 1.3305693864822388,
      "learning_rate": 4.5e-05,
      "loss": 0.1359,
      "step": 960
    },
    {
      "epoch": 10.0,
      "eval_LOCATION_f1": 0.8229166666666666,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.797979797979798,
      "eval_LOCATION_recall": 0.8494623655913979,
      "eval_ORGANIZATION_f1": 0.8011363636363636,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.7580645161290323,
      "eval_ORGANIZATION_recall": 0.8493975903614458,
      "eval_PERSON_f1": 0.9415807560137458,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9194630872483222,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.09369462728500366,
      "eval_overall_accuracy": 0.9711934156378601,
      "eval_overall_f1": 0.8550898203592814,
      "eval_overall_precision": 0.8225806451612904,
      "eval_overall_recall": 0.8902743142144638,
      "eval_runtime": 0.2851,
      "eval_samples_per_second": 596.337,
      "eval_steps_per_second": 10.524,
      "step": 960
    },
    {
      "epoch": 11.0,
      "grad_norm": 0.8002439141273499,
      "learning_rate": 4.4500000000000004e-05,
      "loss": 0.126,
      "step": 1056
    },
    {
      "epoch": 11.0,
      "eval_LOCATION_f1": 0.8205128205128205,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.7843137254901961,
      "eval_LOCATION_recall": 0.8602150537634409,
      "eval_ORGANIZATION_f1": 0.8171428571428571,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.7771739130434783,
      "eval_ORGANIZATION_recall": 0.8614457831325302,
      "eval_PERSON_f1": 0.9383561643835617,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9133333333333333,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.08726800233125687,
      "eval_overall_accuracy": 0.9725651577503429,
      "eval_overall_f1": 0.860215053763441,
      "eval_overall_precision": 0.8256880733944955,
      "eval_overall_recall": 0.8977556109725686,
      "eval_runtime": 0.2848,
      "eval_samples_per_second": 596.875,
      "eval_steps_per_second": 10.533,
      "step": 1056
    },
    {
      "epoch": 12.0,
      "grad_norm": 5.6661248207092285,
      "learning_rate": 4.4000000000000006e-05,
      "loss": 0.1191,
      "step": 1152
    },
    {
      "epoch": 12.0,
      "eval_LOCATION_f1": 0.83248730964467,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.7884615384615384,
      "eval_LOCATION_recall": 0.8817204301075269,
      "eval_ORGANIZATION_f1": 0.8328611898016998,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.786096256684492,
      "eval_ORGANIZATION_recall": 0.8855421686746988,
      "eval_PERSON_f1": 0.9415807560137458,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9194630872483222,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.08264537155628204,
      "eval_overall_accuracy": 0.9739368998628258,
      "eval_overall_f1": 0.8703923900118906,
      "eval_overall_precision": 0.8318181818181818,
      "eval_overall_recall": 0.912718204488778,
      "eval_runtime": 0.2805,
      "eval_samples_per_second": 605.976,
      "eval_steps_per_second": 10.694,
      "step": 1152
    },
    {
      "epoch": 13.0,
      "grad_norm": 1.2274419069290161,
      "learning_rate": 4.35e-05,
      "loss": 0.1126,
      "step": 1248
    },
    {
      "epoch": 13.0,
      "eval_LOCATION_f1": 0.8615384615384616,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8235294117647058,
      "eval_LOCATION_recall": 0.9032258064516129,
      "eval_ORGANIZATION_f1": 0.8497109826589595,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8166666666666667,
      "eval_ORGANIZATION_recall": 0.8855421686746988,
      "eval_PERSON_f1": 0.9480968858131488,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9319727891156463,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.07422925531864166,
      "eval_overall_accuracy": 0.9769547325102881,
      "eval_overall_f1": 0.8867469879518073,
      "eval_overall_precision": 0.8578088578088578,
      "eval_overall_recall": 0.9177057356608479,
      "eval_runtime": 0.2855,
      "eval_samples_per_second": 595.43,
      "eval_steps_per_second": 10.508,
      "step": 1248
    },
    {
      "epoch": 14.0,
      "grad_norm": 1.3920135498046875,
      "learning_rate": 4.3e-05,
      "loss": 0.1061,
      "step": 1344
    },
    {
      "epoch": 14.0,
      "eval_LOCATION_f1": 0.8808290155440415,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.85,
      "eval_LOCATION_recall": 0.9139784946236559,
      "eval_ORGANIZATION_f1": 0.8613569321533923,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8439306358381503,
      "eval_ORGANIZATION_recall": 0.8795180722891566,
      "eval_PERSON_f1": 0.9480968858131488,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9319727891156463,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.07068616151809692,
      "eval_overall_accuracy": 0.978875171467764,
      "eval_overall_f1": 0.8964677222898904,
      "eval_overall_precision": 0.8761904761904762,
      "eval_overall_recall": 0.9177057356608479,
      "eval_runtime": 0.2797,
      "eval_samples_per_second": 607.712,
      "eval_steps_per_second": 10.724,
      "step": 1344
    },
    {
      "epoch": 15.0,
      "grad_norm": 0.7759754061698914,
      "learning_rate": 4.25e-05,
      "loss": 0.1003,
      "step": 1440
    },
    {
      "epoch": 15.0,
      "eval_LOCATION_f1": 0.8911917098445595,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.86,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.8612716763005781,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8277777777777777,
      "eval_ORGANIZATION_recall": 0.8975903614457831,
      "eval_PERSON_f1": 0.9547038327526133,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9448275862068966,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.0703127458691597,
      "eval_overall_accuracy": 0.9783264746227709,
      "eval_overall_f1": 0.9007263922518159,
      "eval_overall_precision": 0.8752941176470588,
      "eval_overall_recall": 0.9276807980049875,
      "eval_runtime": 0.2858,
      "eval_samples_per_second": 594.799,
      "eval_steps_per_second": 10.496,
      "step": 1440
    },
    {
      "epoch": 16.0,
      "grad_norm": 1.396679401397705,
      "learning_rate": 4.2e-05,
      "loss": 0.1008,
      "step": 1536
    },
    {
      "epoch": 16.0,
      "eval_LOCATION_f1": 0.8923076923076922,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8529411764705882,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.8645533141210374,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8287292817679558,
      "eval_ORGANIZATION_recall": 0.9036144578313253,
      "eval_PERSON_f1": 0.9480968858131488,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9319727891156463,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.06862790882587433,
      "eval_overall_accuracy": 0.9777777777777777,
      "eval_overall_f1": 0.9001203369434417,
      "eval_overall_precision": 0.8697674418604651,
      "eval_overall_recall": 0.9326683291770573,
      "eval_runtime": 0.2841,
      "eval_samples_per_second": 598.386,
      "eval_steps_per_second": 10.56,
      "step": 1536
    },
    {
      "epoch": 17.0,
      "grad_norm": 0.9365902543067932,
      "learning_rate": 4.15e-05,
      "loss": 0.0957,
      "step": 1632
    },
    {
      "epoch": 17.0,
      "eval_LOCATION_f1": 0.8911917098445595,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.86,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.8790560471976401,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.861271676300578,
      "eval_ORGANIZATION_recall": 0.8975903614457831,
      "eval_PERSON_f1": 0.9580419580419581,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9513888888888888,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.061672553420066833,
      "eval_overall_accuracy": 0.980246913580247,
      "eval_overall_f1": 0.9095354523227384,
      "eval_overall_precision": 0.8920863309352518,
      "eval_overall_recall": 0.9276807980049875,
      "eval_runtime": 0.2841,
      "eval_samples_per_second": 598.364,
      "eval_steps_per_second": 10.559,
      "step": 1632
    },
    {
      "epoch": 18.0,
      "grad_norm": 0.5375998616218567,
      "learning_rate": 4.1e-05,
      "loss": 0.0923,
      "step": 1728
    },
    {
      "epoch": 18.0,
      "eval_LOCATION_f1": 0.8958333333333334,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8686868686868687,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.8842729970326408,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8713450292397661,
      "eval_ORGANIZATION_recall": 0.8975903614457831,
      "eval_PERSON_f1": 0.9614035087719298,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.958041958041958,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.05938532203435898,
      "eval_overall_accuracy": 0.9818930041152263,
      "eval_overall_f1": 0.9140049140049139,
      "eval_overall_precision": 0.9007263922518159,
      "eval_overall_recall": 0.9276807980049875,
      "eval_runtime": 0.2789,
      "eval_samples_per_second": 609.62,
      "eval_steps_per_second": 10.758,
      "step": 1728
    },
    {
      "epoch": 19.0,
      "grad_norm": 0.978545069694519,
      "learning_rate": 4.05e-05,
      "loss": 0.0894,
      "step": 1824
    },
    {
      "epoch": 19.0,
      "eval_LOCATION_f1": 0.8923076923076922,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8529411764705882,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.8672566371681416,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8497109826589595,
      "eval_ORGANIZATION_recall": 0.8855421686746988,
      "eval_PERSON_f1": 0.9547038327526133,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9448275862068966,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.05911582335829735,
      "eval_overall_accuracy": 0.9799725651577503,
      "eval_overall_f1": 0.9037758830694275,
      "eval_overall_precision": 0.8833333333333333,
      "eval_overall_recall": 0.9251870324189526,
      "eval_runtime": 0.2842,
      "eval_samples_per_second": 598.194,
      "eval_steps_per_second": 10.556,
      "step": 1824
    },
    {
      "epoch": 20.0,
      "grad_norm": 1.320235252380371,
      "learning_rate": 4e-05,
      "loss": 0.0852,
      "step": 1920
    },
    {
      "epoch": 20.0,
      "eval_LOCATION_f1": 0.883248730964467,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8365384615384616,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.8742514970059881,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8690476190476191,
      "eval_ORGANIZATION_recall": 0.8795180722891566,
      "eval_PERSON_f1": 0.9547038327526133,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9448275862068966,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.05648750439286232,
      "eval_overall_accuracy": 0.9813443072702333,
      "eval_overall_f1": 0.9046454767726162,
      "eval_overall_precision": 0.8872901678657075,
      "eval_overall_recall": 0.9226932668329177,
      "eval_runtime": 0.2816,
      "eval_samples_per_second": 603.678,
      "eval_steps_per_second": 10.653,
      "step": 1920
    },
    {
      "epoch": 21.0,
      "grad_norm": 1.0660918951034546,
      "learning_rate": 3.9500000000000005e-05,
      "loss": 0.0857,
      "step": 2016
    },
    {
      "epoch": 21.0,
      "eval_LOCATION_f1": 0.8787878787878788,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8285714285714286,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.8739002932551319,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8514285714285714,
      "eval_ORGANIZATION_recall": 0.8975903614457831,
      "eval_PERSON_f1": 0.9614035087719298,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.958041958041958,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.059097353368997574,
      "eval_overall_accuracy": 0.9816186556927298,
      "eval_overall_f1": 0.9053398058252428,
      "eval_overall_precision": 0.8817966903073287,
      "eval_overall_recall": 0.9301745635910225,
      "eval_runtime": 0.2823,
      "eval_samples_per_second": 602.231,
      "eval_steps_per_second": 10.628,
      "step": 2016
    },
    {
      "epoch": 22.0,
      "grad_norm": 1.226671576499939,
      "learning_rate": 3.9000000000000006e-05,
      "loss": 0.0817,
      "step": 2112
    },
    {
      "epoch": 22.0,
      "eval_LOCATION_f1": 0.8787878787878788,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8285714285714286,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.8705882352941176,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8505747126436781,
      "eval_ORGANIZATION_recall": 0.891566265060241,
      "eval_PERSON_f1": 0.9547038327526133,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9448275862068966,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.05854226276278496,
      "eval_overall_accuracy": 0.9807956104252401,
      "eval_overall_f1": 0.9018181818181819,
      "eval_overall_precision": 0.8773584905660378,
      "eval_overall_recall": 0.9276807980049875,
      "eval_runtime": 0.2824,
      "eval_samples_per_second": 601.972,
      "eval_steps_per_second": 10.623,
      "step": 2112
    },
    {
      "epoch": 23.0,
      "grad_norm": 1.1954342126846313,
      "learning_rate": 3.85e-05,
      "loss": 0.0792,
      "step": 2208
    },
    {
      "epoch": 23.0,
      "eval_LOCATION_f1": 0.882051282051282,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8431372549019608,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.8674698795180723,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8674698795180723,
      "eval_ORGANIZATION_recall": 0.8674698795180723,
      "eval_PERSON_f1": 0.9614035087719298,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.958041958041958,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.05442183464765549,
      "eval_overall_accuracy": 0.9810699588477366,
      "eval_overall_f1": 0.9039408866995073,
      "eval_overall_precision": 0.8929440389294404,
      "eval_overall_recall": 0.9152119700748129,
      "eval_runtime": 0.2813,
      "eval_samples_per_second": 604.426,
      "eval_steps_per_second": 10.666,
      "step": 2208
    },
    {
      "epoch": 24.0,
      "grad_norm": 1.1046844720840454,
      "learning_rate": 3.8e-05,
      "loss": 0.0788,
      "step": 2304
    },
    {
      "epoch": 24.0,
      "eval_LOCATION_f1": 0.8730964467005077,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8269230769230769,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.8674698795180723,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8674698795180723,
      "eval_ORGANIZATION_recall": 0.8674698795180723,
      "eval_PERSON_f1": 0.9614035087719298,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.958041958041958,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.05480505898594856,
      "eval_overall_accuracy": 0.9810699588477366,
      "eval_overall_f1": 0.9017199017199016,
      "eval_overall_precision": 0.8886198547215496,
      "eval_overall_recall": 0.9152119700748129,
      "eval_runtime": 0.2815,
      "eval_samples_per_second": 603.885,
      "eval_steps_per_second": 10.657,
      "step": 2304
    },
    {
      "epoch": 25.0,
      "grad_norm": 1.003726840019226,
      "learning_rate": 3.7500000000000003e-05,
      "loss": 0.0772,
      "step": 2400
    },
    {
      "epoch": 25.0,
      "eval_LOCATION_f1": 0.883248730964467,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8365384615384616,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.8802395209580839,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.875,
      "eval_ORGANIZATION_recall": 0.8855421686746988,
      "eval_PERSON_f1": 0.9580419580419581,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9513888888888888,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.05413680523633957,
      "eval_overall_accuracy": 0.9816186556927298,
      "eval_overall_f1": 0.9082007343941249,
      "eval_overall_precision": 0.8918269230769231,
      "eval_overall_recall": 0.9251870324189526,
      "eval_runtime": 0.285,
      "eval_samples_per_second": 596.532,
      "eval_steps_per_second": 10.527,
      "step": 2400
    },
    {
      "epoch": 26.0,
      "grad_norm": 0.39463770389556885,
      "learning_rate": 3.7e-05,
      "loss": 0.0755,
      "step": 2496
    },
    {
      "epoch": 26.0,
      "eval_LOCATION_f1": 0.900523560209424,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8775510204081632,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.890207715133531,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8771929824561403,
      "eval_ORGANIZATION_recall": 0.9036144578313253,
      "eval_PERSON_f1": 0.9580419580419581,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.9513888888888888,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.05070779472589493,
      "eval_overall_accuracy": 0.9835390946502057,
      "eval_overall_f1": 0.9164619164619164,
      "eval_overall_precision": 0.9031476997578692,
      "eval_overall_recall": 0.9301745635910225,
      "eval_runtime": 0.2828,
      "eval_samples_per_second": 601.218,
      "eval_steps_per_second": 10.61,
      "step": 2496
    },
    {
      "epoch": 27.0,
      "grad_norm": 2.091871500015259,
      "learning_rate": 3.65e-05,
      "loss": 0.0717,
      "step": 2592
    },
    {
      "epoch": 27.0,
      "eval_LOCATION_f1": 0.8958333333333334,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8686868686868687,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.8882352941176471,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.867816091954023,
      "eval_ORGANIZATION_recall": 0.9096385542168675,
      "eval_PERSON_f1": 0.9614035087719298,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.958041958041958,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.05057214945554733,
      "eval_overall_accuracy": 0.9840877914951989,
      "eval_overall_f1": 0.9155446756425949,
      "eval_overall_precision": 0.8990384615384616,
      "eval_overall_recall": 0.9326683291770573,
      "eval_runtime": 0.2821,
      "eval_samples_per_second": 602.66,
      "eval_steps_per_second": 10.635,
      "step": 2592
    },
    {
      "epoch": 28.0,
      "grad_norm": 1.2802491188049316,
      "learning_rate": 3.6e-05,
      "loss": 0.0725,
      "step": 2688
    },
    {
      "epoch": 28.0,
      "eval_LOCATION_f1": 0.8775510204081631,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8349514563106796,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.8869047619047619,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8764705882352941,
      "eval_ORGANIZATION_recall": 0.8975903614457831,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.05177851393818855,
      "eval_overall_accuracy": 0.9832647462277092,
      "eval_overall_f1": 0.9142156862745099,
      "eval_overall_precision": 0.8987951807228916,
      "eval_overall_recall": 0.9301745635910225,
      "eval_runtime": 0.284,
      "eval_samples_per_second": 598.592,
      "eval_steps_per_second": 10.563,
      "step": 2688
    },
    {
      "epoch": 29.0,
      "grad_norm": 0.523560643196106,
      "learning_rate": 3.55e-05,
      "loss": 0.0713,
      "step": 2784
    },
    {
      "epoch": 29.0,
      "eval_LOCATION_f1": 0.882051282051282,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8431372549019608,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.88955223880597,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8816568047337278,
      "eval_ORGANIZATION_recall": 0.8975903614457831,
      "eval_PERSON_f1": 0.9614035087719298,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.958041958041958,
      "eval_PERSON_recall": 0.9647887323943662,
      "eval_loss": 0.050521016120910645,
      "eval_overall_accuracy": 0.9832647462277092,
      "eval_overall_f1": 0.9128834355828221,
      "eval_overall_precision": 0.8985507246376812,
      "eval_overall_recall": 0.9276807980049875,
      "eval_runtime": 0.2811,
      "eval_samples_per_second": 604.758,
      "eval_steps_per_second": 10.672,
      "step": 2784
    },
    {
      "epoch": 30.0,
      "grad_norm": 1.3990178108215332,
      "learning_rate": 3.5e-05,
      "loss": 0.0671,
      "step": 2880
    },
    {
      "epoch": 30.0,
      "eval_LOCATION_f1": 0.8958333333333334,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8686868686868687,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.9020771513353115,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8888888888888888,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.047653622925281525,
      "eval_overall_accuracy": 0.984636488340192,
      "eval_overall_f1": 0.9249692496924969,
      "eval_overall_precision": 0.912621359223301,
      "eval_overall_recall": 0.9376558603491272,
      "eval_runtime": 0.2809,
      "eval_samples_per_second": 605.116,
      "eval_steps_per_second": 10.679,
      "step": 2880
    },
    {
      "epoch": 31.0,
      "grad_norm": 0.28008055686950684,
      "learning_rate": 3.45e-05,
      "loss": 0.0666,
      "step": 2976
    },
    {
      "epoch": 31.0,
      "eval_LOCATION_f1": 0.8775510204081631,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8349514563106796,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.8855421686746989,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8855421686746988,
      "eval_ORGANIZATION_recall": 0.8855421686746988,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04800420626997948,
      "eval_overall_accuracy": 0.9838134430727024,
      "eval_overall_f1": 0.9137931034482758,
      "eval_overall_precision": 0.902676399026764,
      "eval_overall_recall": 0.9251870324189526,
      "eval_runtime": 0.2841,
      "eval_samples_per_second": 598.335,
      "eval_steps_per_second": 10.559,
      "step": 2976
    },
    {
      "epoch": 32.0,
      "grad_norm": 4.019622325897217,
      "learning_rate": 3.4000000000000007e-05,
      "loss": 0.0638,
      "step": 3072
    },
    {
      "epoch": 32.0,
      "eval_LOCATION_f1": 0.8865979381443299,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8514851485148515,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.8948948948948949,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8922155688622755,
      "eval_ORGANIZATION_recall": 0.8975903614457831,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.048197515308856964,
      "eval_overall_accuracy": 0.9843621399176955,
      "eval_overall_f1": 0.9198520345252775,
      "eval_overall_precision": 0.9097560975609756,
      "eval_overall_recall": 0.9301745635910225,
      "eval_runtime": 0.2816,
      "eval_samples_per_second": 603.772,
      "eval_steps_per_second": 10.655,
      "step": 3072
    },
    {
      "epoch": 33.0,
      "grad_norm": 0.4553819000720978,
      "learning_rate": 3.35e-05,
      "loss": 0.0647,
      "step": 3168
    },
    {
      "epoch": 33.0,
      "eval_LOCATION_f1": 0.8775510204081631,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8349514563106796,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.888888888888889,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8862275449101796,
      "eval_ORGANIZATION_recall": 0.891566265060241,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.048246119171381,
      "eval_overall_accuracy": 0.9835390946502057,
      "eval_overall_f1": 0.915129151291513,
      "eval_overall_precision": 0.9029126213592233,
      "eval_overall_recall": 0.9276807980049875,
      "eval_runtime": 0.2812,
      "eval_samples_per_second": 604.507,
      "eval_steps_per_second": 10.668,
      "step": 3168
    },
    {
      "epoch": 34.0,
      "grad_norm": 0.8945901393890381,
      "learning_rate": 3.3e-05,
      "loss": 0.0642,
      "step": 3264
    },
    {
      "epoch": 34.0,
      "eval_LOCATION_f1": 0.882051282051282,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8431372549019608,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.8934911242603549,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.877906976744186,
      "eval_ORGANIZATION_recall": 0.9096385542168675,
      "eval_PERSON_f1": 0.968421052631579,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.965034965034965,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.048569679260253906,
      "eval_overall_accuracy": 0.9832647462277092,
      "eval_overall_f1": 0.9168704156479217,
      "eval_overall_precision": 0.8992805755395683,
      "eval_overall_recall": 0.9351620947630923,
      "eval_runtime": 0.2909,
      "eval_samples_per_second": 584.352,
      "eval_steps_per_second": 10.312,
      "step": 3264
    },
    {
      "epoch": 35.0,
      "grad_norm": 1.6167196035385132,
      "learning_rate": 3.2500000000000004e-05,
      "loss": 0.0603,
      "step": 3360
    },
    {
      "epoch": 35.0,
      "eval_LOCATION_f1": 0.8865979381443299,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8514851485148515,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.8982035928143712,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8928571428571429,
      "eval_ORGANIZATION_recall": 0.9036144578313253,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04634106531739235,
      "eval_overall_accuracy": 0.9851851851851852,
      "eval_overall_f1": 0.9211822660098522,
      "eval_overall_precision": 0.9099756690997567,
      "eval_overall_recall": 0.9326683291770573,
      "eval_runtime": 0.2828,
      "eval_samples_per_second": 601.2,
      "eval_steps_per_second": 10.609,
      "step": 3360
    },
    {
      "epoch": 36.0,
      "grad_norm": 0.5620716214179993,
      "learning_rate": 3.2000000000000005e-05,
      "loss": 0.0627,
      "step": 3456
    },
    {
      "epoch": 36.0,
      "eval_LOCATION_f1": 0.8775510204081631,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8349514563106796,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.8955223880597014,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8875739644970414,
      "eval_ORGANIZATION_recall": 0.9036144578313253,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04827764630317688,
      "eval_overall_accuracy": 0.984636488340192,
      "eval_overall_f1": 0.9177914110429448,
      "eval_overall_precision": 0.9033816425120773,
      "eval_overall_recall": 0.9326683291770573,
      "eval_runtime": 0.2836,
      "eval_samples_per_second": 599.34,
      "eval_steps_per_second": 10.577,
      "step": 3456
    },
    {
      "epoch": 37.0,
      "grad_norm": 2.3784847259521484,
      "learning_rate": 3.15e-05,
      "loss": 0.0606,
      "step": 3552
    },
    {
      "epoch": 37.0,
      "eval_LOCATION_f1": 0.900523560209424,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8775510204081632,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.9085545722713864,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8901734104046243,
      "eval_ORGANIZATION_recall": 0.927710843373494,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04605403542518616,
      "eval_overall_accuracy": 0.9854595336076818,
      "eval_overall_f1": 0.9287469287469288,
      "eval_overall_precision": 0.9152542372881356,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.287,
      "eval_samples_per_second": 592.388,
      "eval_steps_per_second": 10.454,
      "step": 3552
    },
    {
      "epoch": 38.0,
      "grad_norm": 0.48580947518348694,
      "learning_rate": 3.1e-05,
      "loss": 0.0602,
      "step": 3648
    },
    {
      "epoch": 38.0,
      "eval_LOCATION_f1": 0.9100529100529101,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8958333333333334,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.9112426035502958,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8953488372093024,
      "eval_ORGANIZATION_recall": 0.927710843373494,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04565339908003807,
      "eval_overall_accuracy": 0.9857338820301783,
      "eval_overall_f1": 0.9321824907521579,
      "eval_overall_precision": 0.9219512195121952,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2829,
      "eval_samples_per_second": 601.012,
      "eval_steps_per_second": 10.606,
      "step": 3648
    },
    {
      "epoch": 39.0,
      "grad_norm": 1.0675240755081177,
      "learning_rate": 3.05e-05,
      "loss": 0.058,
      "step": 3744
    },
    {
      "epoch": 39.0,
      "eval_LOCATION_f1": 0.9052631578947367,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8865979381443299,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.9085545722713864,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8901734104046243,
      "eval_ORGANIZATION_recall": 0.927710843373494,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04524984955787659,
      "eval_overall_accuracy": 0.9860082304526749,
      "eval_overall_f1": 0.929889298892989,
      "eval_overall_precision": 0.9174757281553398,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2837,
      "eval_samples_per_second": 599.125,
      "eval_steps_per_second": 10.573,
      "step": 3744
    },
    {
      "epoch": 40.0,
      "grad_norm": 1.1679747104644775,
      "learning_rate": 3e-05,
      "loss": 0.0579,
      "step": 3840
    },
    {
      "epoch": 40.0,
      "eval_LOCATION_f1": 0.9100529100529101,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8958333333333334,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.9107142857142857,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.044337525963783264,
      "eval_overall_accuracy": 0.9862825788751715,
      "eval_overall_f1": 0.9320148331273177,
      "eval_overall_precision": 0.9240196078431373,
      "eval_overall_recall": 0.940149625935162,
      "eval_runtime": 0.2853,
      "eval_samples_per_second": 595.872,
      "eval_steps_per_second": 10.515,
      "step": 3840
    },
    {
      "epoch": 41.0,
      "grad_norm": 0.8956630229949951,
      "learning_rate": 2.95e-05,
      "loss": 0.0551,
      "step": 3936
    },
    {
      "epoch": 41.0,
      "eval_LOCATION_f1": 0.9100529100529101,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8958333333333334,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.9144542772861357,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8959537572254336,
      "eval_ORGANIZATION_recall": 0.9337349397590361,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04385361447930336,
      "eval_overall_accuracy": 0.9868312757201646,
      "eval_overall_f1": 0.9334975369458128,
      "eval_overall_precision": 0.9221411192214112,
      "eval_overall_recall": 0.9451371571072319,
      "eval_runtime": 0.2844,
      "eval_samples_per_second": 597.724,
      "eval_steps_per_second": 10.548,
      "step": 3936
    },
    {
      "epoch": 42.0,
      "grad_norm": 1.3600409030914307,
      "learning_rate": 2.9e-05,
      "loss": 0.0568,
      "step": 4032
    },
    {
      "epoch": 42.0,
      "eval_LOCATION_f1": 0.90625,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8787878787878788,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9047619047619048,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8941176470588236,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.043516892939805984,
      "eval_overall_accuracy": 0.9865569272976681,
      "eval_overall_f1": 0.9285714285714285,
      "eval_overall_precision": 0.9172749391727494,
      "eval_overall_recall": 0.940149625935162,
      "eval_runtime": 0.2806,
      "eval_samples_per_second": 605.74,
      "eval_steps_per_second": 10.69,
      "step": 4032
    },
    {
      "epoch": 43.0,
      "grad_norm": 0.8620499968528748,
      "learning_rate": 2.8499999999999998e-05,
      "loss": 0.0557,
      "step": 4128
    },
    {
      "epoch": 43.0,
      "eval_LOCATION_f1": 0.9157894736842105,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8969072164948454,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.906906906906907,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9041916167664671,
      "eval_ORGANIZATION_recall": 0.9096385542168675,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04396757110953331,
      "eval_overall_accuracy": 0.9860082304526749,
      "eval_overall_f1": 0.9318463444857497,
      "eval_overall_precision": 0.9261083743842364,
      "eval_overall_recall": 0.9376558603491272,
      "eval_runtime": 0.2853,
      "eval_samples_per_second": 595.869,
      "eval_steps_per_second": 10.515,
      "step": 4128
    },
    {
      "epoch": 44.0,
      "grad_norm": 0.3250766396522522,
      "learning_rate": 2.8000000000000003e-05,
      "loss": 0.0582,
      "step": 4224
    },
    {
      "epoch": 44.0,
      "eval_LOCATION_f1": 0.8923076923076922,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8529411764705882,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.896969696969697,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9024390243902439,
      "eval_ORGANIZATION_recall": 0.891566265060241,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04460552707314491,
      "eval_overall_accuracy": 0.9843621399176955,
      "eval_overall_f1": 0.9221260815822002,
      "eval_overall_precision": 0.9142156862745098,
      "eval_overall_recall": 0.9301745635910225,
      "eval_runtime": 0.2824,
      "eval_samples_per_second": 601.921,
      "eval_steps_per_second": 10.622,
      "step": 4224
    },
    {
      "epoch": 45.0,
      "grad_norm": 1.508424997329712,
      "learning_rate": 2.7500000000000004e-05,
      "loss": 0.0548,
      "step": 4320
    },
    {
      "epoch": 45.0,
      "eval_LOCATION_f1": 0.9109947643979057,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8877551020408163,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9161676646706587,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9107142857142857,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.042420320212841034,
      "eval_overall_accuracy": 0.9865569272976681,
      "eval_overall_f1": 0.9344870210135972,
      "eval_overall_precision": 0.9264705882352942,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2838,
      "eval_samples_per_second": 598.985,
      "eval_steps_per_second": 10.57,
      "step": 4320
    },
    {
      "epoch": 46.0,
      "grad_norm": 0.7623454928398132,
      "learning_rate": 2.7000000000000002e-05,
      "loss": 0.0533,
      "step": 4416
    },
    {
      "epoch": 46.0,
      "eval_LOCATION_f1": 0.90625,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8787878787878788,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9041916167664671,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8988095238095238,
      "eval_ORGANIZATION_recall": 0.9096385542168675,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.0423850379884243,
      "eval_overall_accuracy": 0.9865569272976681,
      "eval_overall_f1": 0.928395061728395,
      "eval_overall_precision": 0.9193154034229829,
      "eval_overall_recall": 0.9376558603491272,
      "eval_runtime": 0.2812,
      "eval_samples_per_second": 604.565,
      "eval_steps_per_second": 10.669,
      "step": 4416
    },
    {
      "epoch": 47.0,
      "grad_norm": 0.9061383008956909,
      "learning_rate": 2.6500000000000004e-05,
      "loss": 0.0516,
      "step": 4512
    },
    {
      "epoch": 47.0,
      "eval_LOCATION_f1": 0.8958333333333334,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8686868686868687,
      "eval_LOCATION_recall": 0.9247311827956989,
      "eval_ORGANIZATION_f1": 0.912280701754386,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8863636363636364,
      "eval_ORGANIZATION_recall": 0.9397590361445783,
      "eval_PERSON_f1": 0.968421052631579,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.965034965034965,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.0428488552570343,
      "eval_overall_accuracy": 0.9860082304526749,
      "eval_overall_f1": 0.9279609279609279,
      "eval_overall_precision": 0.9090909090909091,
      "eval_overall_recall": 0.9476309226932669,
      "eval_runtime": 0.283,
      "eval_samples_per_second": 600.703,
      "eval_steps_per_second": 10.601,
      "step": 4512
    },
    {
      "epoch": 48.0,
      "grad_norm": 0.753669798374176,
      "learning_rate": 2.6000000000000002e-05,
      "loss": 0.0501,
      "step": 4608
    },
    {
      "epoch": 48.0,
      "eval_LOCATION_f1": 0.90625,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8787878787878788,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.906906906906907,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9041916167664671,
      "eval_ORGANIZATION_recall": 0.9096385542168675,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04297007992863655,
      "eval_overall_accuracy": 0.9862825788751715,
      "eval_overall_f1": 0.9295426452410381,
      "eval_overall_precision": 0.9215686274509803,
      "eval_overall_recall": 0.9376558603491272,
      "eval_runtime": 0.2827,
      "eval_samples_per_second": 601.339,
      "eval_steps_per_second": 10.612,
      "step": 4608
    },
    {
      "epoch": 49.0,
      "grad_norm": 1.0571324825286865,
      "learning_rate": 2.5500000000000003e-05,
      "loss": 0.053,
      "step": 4704
    },
    {
      "epoch": 49.0,
      "eval_LOCATION_f1": 0.90625,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8787878787878788,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9134328358208955,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9053254437869822,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04329885542392731,
      "eval_overall_accuracy": 0.9865569272976681,
      "eval_overall_f1": 0.9321824907521579,
      "eval_overall_precision": 0.9219512195121952,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2818,
      "eval_samples_per_second": 603.355,
      "eval_steps_per_second": 10.647,
      "step": 4704
    },
    {
      "epoch": 50.0,
      "grad_norm": 0.5389184355735779,
      "learning_rate": 2.5e-05,
      "loss": 0.0483,
      "step": 4800
    },
    {
      "epoch": 50.0,
      "eval_LOCATION_f1": 0.9206349206349206,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.90625,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04164545238018036,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9343246592317225,
      "eval_overall_precision": 0.9285714285714286,
      "eval_overall_recall": 0.940149625935162,
      "eval_runtime": 0.2839,
      "eval_samples_per_second": 598.843,
      "eval_steps_per_second": 10.568,
      "step": 4800
    },
    {
      "epoch": 51.0,
      "grad_norm": 0.6280408501625061,
      "learning_rate": 2.45e-05,
      "loss": 0.0505,
      "step": 4896
    },
    {
      "epoch": 51.0,
      "eval_LOCATION_f1": 0.9214659685863874,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8979591836734694,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9096385542168675,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9096385542168675,
      "eval_ORGANIZATION_recall": 0.9096385542168675,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04181710258126259,
      "eval_overall_accuracy": 0.9865569272976681,
      "eval_overall_f1": 0.9343246592317225,
      "eval_overall_precision": 0.9285714285714286,
      "eval_overall_recall": 0.940149625935162,
      "eval_runtime": 0.2821,
      "eval_samples_per_second": 602.699,
      "eval_steps_per_second": 10.636,
      "step": 4896
    },
    {
      "epoch": 52.0,
      "grad_norm": 0.9616138935089111,
      "learning_rate": 2.4e-05,
      "loss": 0.05,
      "step": 4992
    },
    {
      "epoch": 52.0,
      "eval_LOCATION_f1": 0.9304812834224598,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.925531914893617,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9053254437869822,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8895348837209303,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04034818708896637,
      "eval_overall_accuracy": 0.9879286694101509,
      "eval_overall_f1": 0.9344870210135972,
      "eval_overall_precision": 0.9264705882352942,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2816,
      "eval_samples_per_second": 603.695,
      "eval_steps_per_second": 10.653,
      "step": 4992
    },
    {
      "epoch": 53.0,
      "grad_norm": 0.9302768111228943,
      "learning_rate": 2.35e-05,
      "loss": 0.0493,
      "step": 5088
    },
    {
      "epoch": 53.0,
      "eval_LOCATION_f1": 0.9157894736842105,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8969072164948454,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04219473525881767,
      "eval_overall_accuracy": 0.9860082304526749,
      "eval_overall_f1": 0.9331683168316831,
      "eval_overall_precision": 0.9262899262899262,
      "eval_overall_recall": 0.940149625935162,
      "eval_runtime": 0.2812,
      "eval_samples_per_second": 604.63,
      "eval_steps_per_second": 10.67,
      "step": 5088
    },
    {
      "epoch": 54.0,
      "grad_norm": 0.40551334619522095,
      "learning_rate": 2.3000000000000003e-05,
      "loss": 0.0487,
      "step": 5184
    },
    {
      "epoch": 54.0,
      "eval_LOCATION_f1": 0.925531914893617,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.9157894736842105,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9134328358208955,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9053254437869822,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04084800183773041,
      "eval_overall_accuracy": 0.9876543209876543,
      "eval_overall_f1": 0.9368029739776952,
      "eval_overall_precision": 0.9310344827586207,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2813,
      "eval_samples_per_second": 604.233,
      "eval_steps_per_second": 10.663,
      "step": 5184
    },
    {
      "epoch": 55.0,
      "grad_norm": 1.221318244934082,
      "learning_rate": 2.25e-05,
      "loss": 0.0485,
      "step": 5280
    },
    {
      "epoch": 55.0,
      "eval_LOCATION_f1": 0.925531914893617,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.9157894736842105,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9194029850746269,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9112426035502958,
      "eval_ORGANIZATION_recall": 0.927710843373494,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04019373282790184,
      "eval_overall_accuracy": 0.9873799725651577,
      "eval_overall_f1": 0.9392812887236679,
      "eval_overall_precision": 0.9334975369458128,
      "eval_overall_recall": 0.9451371571072319,
      "eval_runtime": 0.2806,
      "eval_samples_per_second": 605.767,
      "eval_steps_per_second": 10.69,
      "step": 5280
    },
    {
      "epoch": 56.0,
      "grad_norm": 2.138538360595703,
      "learning_rate": 2.2000000000000003e-05,
      "loss": 0.0491,
      "step": 5376
    },
    {
      "epoch": 56.0,
      "eval_LOCATION_f1": 0.9109947643979057,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8877551020408163,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9144542772861357,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8959537572254336,
      "eval_ORGANIZATION_recall": 0.9337349397590361,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.043160635977983475,
      "eval_overall_accuracy": 0.9862825788751715,
      "eval_overall_f1": 0.9336609336609337,
      "eval_overall_precision": 0.9200968523002422,
      "eval_overall_recall": 0.9476309226932669,
      "eval_runtime": 0.2797,
      "eval_samples_per_second": 607.881,
      "eval_steps_per_second": 10.727,
      "step": 5376
    },
    {
      "epoch": 57.0,
      "grad_norm": 0.7274695038795471,
      "learning_rate": 2.15e-05,
      "loss": 0.0495,
      "step": 5472
    },
    {
      "epoch": 57.0,
      "eval_LOCATION_f1": 0.9214659685863874,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8979591836734694,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04091630131006241,
      "eval_overall_accuracy": 0.9865569272976681,
      "eval_overall_f1": 0.9344870210135972,
      "eval_overall_precision": 0.9264705882352942,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2809,
      "eval_samples_per_second": 605.197,
      "eval_steps_per_second": 10.68,
      "step": 5472
    },
    {
      "epoch": 58.0,
      "grad_norm": 0.6043224334716797,
      "learning_rate": 2.1e-05,
      "loss": 0.0495,
      "step": 5568
    },
    {
      "epoch": 58.0,
      "eval_LOCATION_f1": 0.9214659685863874,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8979591836734694,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04245884343981743,
      "eval_overall_accuracy": 0.9865569272976681,
      "eval_overall_f1": 0.9344870210135972,
      "eval_overall_precision": 0.9264705882352942,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2821,
      "eval_samples_per_second": 602.589,
      "eval_steps_per_second": 10.634,
      "step": 5568
    },
    {
      "epoch": 59.0,
      "grad_norm": 0.7443007826805115,
      "learning_rate": 2.05e-05,
      "loss": 0.0462,
      "step": 5664
    },
    {
      "epoch": 59.0,
      "eval_LOCATION_f1": 0.9214659685863874,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8979591836734694,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9129129129129129,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9101796407185628,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04124193638563156,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9356435643564357,
      "eval_overall_precision": 0.9287469287469288,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2806,
      "eval_samples_per_second": 605.911,
      "eval_steps_per_second": 10.693,
      "step": 5664
    },
    {
      "epoch": 60.0,
      "grad_norm": 1.2478013038635254,
      "learning_rate": 2e-05,
      "loss": 0.048,
      "step": 5760
    },
    {
      "epoch": 60.0,
      "eval_LOCATION_f1": 0.9263157894736843,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.9072164948453608,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9156626506024096,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9156626506024096,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.040921688079833984,
      "eval_overall_accuracy": 0.9868312757201646,
      "eval_overall_f1": 0.9379652605459058,
      "eval_overall_precision": 0.9333333333333333,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2812,
      "eval_samples_per_second": 604.536,
      "eval_steps_per_second": 10.668,
      "step": 5760
    },
    {
      "epoch": 61.0,
      "grad_norm": 0.783888578414917,
      "learning_rate": 1.9500000000000003e-05,
      "loss": 0.048,
      "step": 5856
    },
    {
      "epoch": 61.0,
      "eval_LOCATION_f1": 0.9214659685863874,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8979591836734694,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9129129129129129,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9101796407185628,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03963721916079521,
      "eval_overall_accuracy": 0.9879286694101509,
      "eval_overall_f1": 0.9356435643564357,
      "eval_overall_precision": 0.9287469287469288,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2837,
      "eval_samples_per_second": 599.237,
      "eval_steps_per_second": 10.575,
      "step": 5856
    },
    {
      "epoch": 62.0,
      "grad_norm": 1.2428171634674072,
      "learning_rate": 1.9e-05,
      "loss": 0.0461,
      "step": 5952
    },
    {
      "epoch": 62.0,
      "eval_LOCATION_f1": 0.9157894736842105,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8969072164948454,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9014925373134328,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.893491124260355,
      "eval_ORGANIZATION_recall": 0.9096385542168675,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.040256600826978683,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9295426452410381,
      "eval_overall_precision": 0.9215686274509803,
      "eval_overall_recall": 0.9376558603491272,
      "eval_runtime": 0.2805,
      "eval_samples_per_second": 606.008,
      "eval_steps_per_second": 10.694,
      "step": 5952
    },
    {
      "epoch": 63.0,
      "grad_norm": 3.580406427383423,
      "learning_rate": 1.85e-05,
      "loss": 0.0459,
      "step": 6048
    },
    {
      "epoch": 63.0,
      "eval_LOCATION_f1": 0.9166666666666667,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8888888888888888,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9189189189189191,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9161676646706587,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04053889960050583,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9369592088998765,
      "eval_overall_precision": 0.928921568627451,
      "eval_overall_recall": 0.9451371571072319,
      "eval_runtime": 0.2795,
      "eval_samples_per_second": 608.325,
      "eval_steps_per_second": 10.735,
      "step": 6048
    },
    {
      "epoch": 64.0,
      "grad_norm": 0.41149088740348816,
      "learning_rate": 1.8e-05,
      "loss": 0.0461,
      "step": 6144
    },
    {
      "epoch": 64.0,
      "eval_LOCATION_f1": 0.9157894736842105,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8969072164948454,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.8988095238095238,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.888235294117647,
      "eval_ORGANIZATION_recall": 0.9096385542168675,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03941579535603523,
      "eval_overall_accuracy": 0.9873799725651577,
      "eval_overall_f1": 0.928395061728395,
      "eval_overall_precision": 0.9193154034229829,
      "eval_overall_recall": 0.9376558603491272,
      "eval_runtime": 0.2836,
      "eval_samples_per_second": 599.391,
      "eval_steps_per_second": 10.577,
      "step": 6144
    },
    {
      "epoch": 65.0,
      "grad_norm": 0.4289607107639313,
      "learning_rate": 1.75e-05,
      "loss": 0.0431,
      "step": 6240
    },
    {
      "epoch": 65.0,
      "eval_LOCATION_f1": 0.9214659685863874,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8979591836734694,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9063444108761329,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9090909090909091,
      "eval_ORGANIZATION_recall": 0.9036144578313253,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.040837135165929794,
      "eval_overall_accuracy": 0.9873799725651577,
      "eval_overall_f1": 0.9330024813895781,
      "eval_overall_precision": 0.928395061728395,
      "eval_overall_recall": 0.9376558603491272,
      "eval_runtime": 0.282,
      "eval_samples_per_second": 602.839,
      "eval_steps_per_second": 10.638,
      "step": 6240
    },
    {
      "epoch": 66.0,
      "grad_norm": 0.7041921615600586,
      "learning_rate": 1.7000000000000003e-05,
      "loss": 0.0448,
      "step": 6336
    },
    {
      "epoch": 66.0,
      "eval_LOCATION_f1": 0.9263157894736843,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.9072164948453608,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9134328358208955,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9053254437869822,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03957487642765045,
      "eval_overall_accuracy": 0.9876543209876543,
      "eval_overall_f1": 0.9369592088998765,
      "eval_overall_precision": 0.928921568627451,
      "eval_overall_recall": 0.9451371571072319,
      "eval_runtime": 0.2803,
      "eval_samples_per_second": 606.538,
      "eval_steps_per_second": 10.704,
      "step": 6336
    },
    {
      "epoch": 67.0,
      "grad_norm": 0.8772794008255005,
      "learning_rate": 1.65e-05,
      "loss": 0.044,
      "step": 6432
    },
    {
      "epoch": 67.0,
      "eval_LOCATION_f1": 0.9214659685863874,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8979591836734694,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9063444108761329,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9090909090909091,
      "eval_ORGANIZATION_recall": 0.9036144578313253,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04030785709619522,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9330024813895781,
      "eval_overall_precision": 0.928395061728395,
      "eval_overall_recall": 0.9376558603491272,
      "eval_runtime": 0.2815,
      "eval_samples_per_second": 603.992,
      "eval_steps_per_second": 10.659,
      "step": 6432
    },
    {
      "epoch": 68.0,
      "grad_norm": 0.5797802209854126,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.0439,
      "step": 6528
    },
    {
      "epoch": 68.0,
      "eval_LOCATION_f1": 0.9166666666666667,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8888888888888888,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9151515151515153,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9207317073170732,
      "eval_ORGANIZATION_recall": 0.9096385542168675,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04043779894709587,
      "eval_overall_accuracy": 0.9873799725651577,
      "eval_overall_f1": 0.9354838709677419,
      "eval_overall_precision": 0.9308641975308642,
      "eval_overall_recall": 0.940149625935162,
      "eval_runtime": 0.2801,
      "eval_samples_per_second": 607.009,
      "eval_steps_per_second": 10.712,
      "step": 6528
    },
    {
      "epoch": 69.0,
      "grad_norm": 1.7521965503692627,
      "learning_rate": 1.55e-05,
      "loss": 0.0451,
      "step": 6624
    },
    {
      "epoch": 69.0,
      "eval_LOCATION_f1": 0.911917098445596,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.88,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9151515151515153,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9207317073170732,
      "eval_ORGANIZATION_recall": 0.9096385542168675,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.0415940061211586,
      "eval_overall_accuracy": 0.9868312757201646,
      "eval_overall_f1": 0.9343246592317225,
      "eval_overall_precision": 0.9285714285714286,
      "eval_overall_recall": 0.940149625935162,
      "eval_runtime": 0.2807,
      "eval_samples_per_second": 605.718,
      "eval_steps_per_second": 10.689,
      "step": 6624
    },
    {
      "epoch": 70.0,
      "grad_norm": 1.2547327280044556,
      "learning_rate": 1.5e-05,
      "loss": 0.0429,
      "step": 6720
    },
    {
      "epoch": 70.0,
      "eval_LOCATION_f1": 0.911917098445596,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.88,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9134328358208955,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9053254437869822,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04034920409321785,
      "eval_overall_accuracy": 0.9876543209876543,
      "eval_overall_f1": 0.9334975369458128,
      "eval_overall_precision": 0.9221411192214112,
      "eval_overall_recall": 0.9451371571072319,
      "eval_runtime": 0.2822,
      "eval_samples_per_second": 602.384,
      "eval_steps_per_second": 10.63,
      "step": 6720
    },
    {
      "epoch": 71.0,
      "grad_norm": 0.7320692539215088,
      "learning_rate": 1.45e-05,
      "loss": 0.0447,
      "step": 6816
    },
    {
      "epoch": 71.0,
      "eval_LOCATION_f1": 0.9214659685863874,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8979591836734694,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9221556886227544,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9166666666666666,
      "eval_ORGANIZATION_recall": 0.927710843373494,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04022010415792465,
      "eval_overall_accuracy": 0.9876543209876543,
      "eval_overall_f1": 0.9394313967861558,
      "eval_overall_precision": 0.9313725490196079,
      "eval_overall_recall": 0.9476309226932669,
      "eval_runtime": 0.2802,
      "eval_samples_per_second": 606.723,
      "eval_steps_per_second": 10.707,
      "step": 6816
    },
    {
      "epoch": 72.0,
      "grad_norm": 0.819017231464386,
      "learning_rate": 1.4000000000000001e-05,
      "loss": 0.0437,
      "step": 6912
    },
    {
      "epoch": 72.0,
      "eval_LOCATION_f1": 0.9166666666666667,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8888888888888888,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9123867069486404,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9151515151515152,
      "eval_ORGANIZATION_recall": 0.9096385542168675,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03983786702156067,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9343246592317225,
      "eval_overall_precision": 0.9285714285714286,
      "eval_overall_recall": 0.940149625935162,
      "eval_runtime": 0.2814,
      "eval_samples_per_second": 604.215,
      "eval_steps_per_second": 10.663,
      "step": 6912
    },
    {
      "epoch": 73.0,
      "grad_norm": 1.5767393112182617,
      "learning_rate": 1.3500000000000001e-05,
      "loss": 0.041,
      "step": 7008
    },
    {
      "epoch": 73.0,
      "eval_LOCATION_f1": 0.9109947643979057,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8877551020408163,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9161676646706587,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9107142857142857,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03988140448927879,
      "eval_overall_accuracy": 0.9873799725651577,
      "eval_overall_f1": 0.9344870210135972,
      "eval_overall_precision": 0.9264705882352942,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2828,
      "eval_samples_per_second": 601.131,
      "eval_steps_per_second": 10.608,
      "step": 7008
    },
    {
      "epoch": 74.0,
      "grad_norm": 0.34249478578567505,
      "learning_rate": 1.3000000000000001e-05,
      "loss": 0.0425,
      "step": 7104
    },
    {
      "epoch": 74.0,
      "eval_LOCATION_f1": 0.9157894736842105,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8969072164948454,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9194029850746269,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9112426035502958,
      "eval_ORGANIZATION_recall": 0.927710843373494,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04064317047595978,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9369592088998765,
      "eval_overall_precision": 0.928921568627451,
      "eval_overall_recall": 0.9451371571072319,
      "eval_runtime": 0.2816,
      "eval_samples_per_second": 603.647,
      "eval_steps_per_second": 10.653,
      "step": 7104
    },
    {
      "epoch": 75.0,
      "grad_norm": 0.588118851184845,
      "learning_rate": 1.25e-05,
      "loss": 0.0426,
      "step": 7200
    },
    {
      "epoch": 75.0,
      "eval_LOCATION_f1": 0.9109947643979057,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8877551020408163,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9107142857142857,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03951849043369293,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9321824907521579,
      "eval_overall_precision": 0.9219512195121952,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2807,
      "eval_samples_per_second": 605.599,
      "eval_steps_per_second": 10.687,
      "step": 7200
    },
    {
      "epoch": 76.0,
      "grad_norm": 0.5939780473709106,
      "learning_rate": 1.2e-05,
      "loss": 0.0398,
      "step": 7296
    },
    {
      "epoch": 76.0,
      "eval_LOCATION_f1": 0.9166666666666667,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8888888888888888,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.918429003021148,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9212121212121213,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.040211234241724014,
      "eval_overall_accuracy": 0.9873799725651577,
      "eval_overall_f1": 0.9368029739776952,
      "eval_overall_precision": 0.9310344827586207,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2831,
      "eval_samples_per_second": 600.543,
      "eval_steps_per_second": 10.598,
      "step": 7296
    },
    {
      "epoch": 77.0,
      "grad_norm": 0.6372460126876831,
      "learning_rate": 1.1500000000000002e-05,
      "loss": 0.0407,
      "step": 7392
    },
    {
      "epoch": 77.0,
      "eval_LOCATION_f1": 0.90625,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8787878787878788,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9096385542168675,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9096385542168675,
      "eval_ORGANIZATION_recall": 0.9096385542168675,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.039185136556625366,
      "eval_overall_accuracy": 0.9873799725651577,
      "eval_overall_f1": 0.9306930693069307,
      "eval_overall_precision": 0.9238329238329238,
      "eval_overall_recall": 0.9376558603491272,
      "eval_runtime": 0.2816,
      "eval_samples_per_second": 603.751,
      "eval_steps_per_second": 10.654,
      "step": 7392
    },
    {
      "epoch": 78.0,
      "grad_norm": 0.7889866828918457,
      "learning_rate": 1.1000000000000001e-05,
      "loss": 0.0411,
      "step": 7488
    },
    {
      "epoch": 78.0,
      "eval_LOCATION_f1": 0.90625,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8787878787878788,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9041916167664671,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8988095238095238,
      "eval_ORGANIZATION_recall": 0.9096385542168675,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03939792886376381,
      "eval_overall_accuracy": 0.9868312757201646,
      "eval_overall_f1": 0.928395061728395,
      "eval_overall_precision": 0.9193154034229829,
      "eval_overall_recall": 0.9376558603491272,
      "eval_runtime": 0.2813,
      "eval_samples_per_second": 604.429,
      "eval_steps_per_second": 10.666,
      "step": 7488
    },
    {
      "epoch": 79.0,
      "grad_norm": 0.8031059503555298,
      "learning_rate": 1.05e-05,
      "loss": 0.0417,
      "step": 7584
    },
    {
      "epoch": 79.0,
      "eval_LOCATION_f1": 0.90625,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8787878787878788,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9134328358208955,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9053254437869822,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.039482057094573975,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9321824907521579,
      "eval_overall_precision": 0.9219512195121952,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2824,
      "eval_samples_per_second": 601.902,
      "eval_steps_per_second": 10.622,
      "step": 7584
    },
    {
      "epoch": 80.0,
      "grad_norm": 2.120938777923584,
      "learning_rate": 1e-05,
      "loss": 0.0412,
      "step": 7680
    },
    {
      "epoch": 80.0,
      "eval_LOCATION_f1": 0.90625,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8787878787878788,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9053254437869822,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8895348837209303,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03960248827934265,
      "eval_overall_accuracy": 0.9876543209876543,
      "eval_overall_f1": 0.9287469287469288,
      "eval_overall_precision": 0.9152542372881356,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2823,
      "eval_samples_per_second": 602.136,
      "eval_steps_per_second": 10.626,
      "step": 7680
    },
    {
      "epoch": 81.0,
      "grad_norm": 0.5607914924621582,
      "learning_rate": 9.5e-06,
      "loss": 0.0431,
      "step": 7776
    },
    {
      "epoch": 81.0,
      "eval_LOCATION_f1": 0.9015544041450777,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.87,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9047619047619048,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8941176470588236,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03992500156164169,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9274292742927428,
      "eval_overall_precision": 0.9150485436893204,
      "eval_overall_recall": 0.940149625935162,
      "eval_runtime": 0.2802,
      "eval_samples_per_second": 606.677,
      "eval_steps_per_second": 10.706,
      "step": 7776
    },
    {
      "epoch": 82.0,
      "grad_norm": 1.140172004699707,
      "learning_rate": 9e-06,
      "loss": 0.042,
      "step": 7872
    },
    {
      "epoch": 82.0,
      "eval_LOCATION_f1": 0.9166666666666667,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8888888888888888,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9189189189189191,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9161676646706587,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04006196931004524,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9369592088998765,
      "eval_overall_precision": 0.928921568627451,
      "eval_overall_recall": 0.9451371571072319,
      "eval_runtime": 0.2823,
      "eval_samples_per_second": 602.241,
      "eval_steps_per_second": 10.628,
      "step": 7872
    },
    {
      "epoch": 83.0,
      "grad_norm": 0.7160885334014893,
      "learning_rate": 8.500000000000002e-06,
      "loss": 0.0412,
      "step": 7968
    },
    {
      "epoch": 83.0,
      "eval_LOCATION_f1": 0.9166666666666667,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8888888888888888,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9189189189189191,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9161676646706587,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.040310610085725784,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9369592088998765,
      "eval_overall_precision": 0.928921568627451,
      "eval_overall_recall": 0.9451371571072319,
      "eval_runtime": 0.2826,
      "eval_samples_per_second": 601.646,
      "eval_steps_per_second": 10.617,
      "step": 7968
    },
    {
      "epoch": 84.0,
      "grad_norm": 0.9255092144012451,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.0413,
      "step": 8064
    },
    {
      "epoch": 84.0,
      "eval_LOCATION_f1": 0.9166666666666667,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8888888888888888,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9129129129129129,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9101796407185628,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.04085477441549301,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9344870210135972,
      "eval_overall_precision": 0.9264705882352942,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2809,
      "eval_samples_per_second": 605.295,
      "eval_steps_per_second": 10.682,
      "step": 8064
    },
    {
      "epoch": 85.0,
      "grad_norm": 0.6887927651405334,
      "learning_rate": 7.5e-06,
      "loss": 0.0405,
      "step": 8160
    },
    {
      "epoch": 85.0,
      "eval_LOCATION_f1": 0.9072164948453608,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8712871287128713,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9161676646706587,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9107142857142857,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03969047591090202,
      "eval_overall_accuracy": 0.9879286694101509,
      "eval_overall_f1": 0.9334975369458128,
      "eval_overall_precision": 0.9221411192214112,
      "eval_overall_recall": 0.9451371571072319,
      "eval_runtime": 0.2815,
      "eval_samples_per_second": 603.924,
      "eval_steps_per_second": 10.657,
      "step": 8160
    },
    {
      "epoch": 86.0,
      "grad_norm": 1.2184239625930786,
      "learning_rate": 7.000000000000001e-06,
      "loss": 0.0405,
      "step": 8256
    },
    {
      "epoch": 86.0,
      "eval_LOCATION_f1": 0.9072164948453608,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8712871287128713,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9161676646706587,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9107142857142857,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03970079496502876,
      "eval_overall_accuracy": 0.9879286694101509,
      "eval_overall_f1": 0.9334975369458128,
      "eval_overall_precision": 0.9221411192214112,
      "eval_overall_recall": 0.9451371571072319,
      "eval_runtime": 0.2832,
      "eval_samples_per_second": 600.222,
      "eval_steps_per_second": 10.592,
      "step": 8256
    },
    {
      "epoch": 87.0,
      "grad_norm": 1.5015846490859985,
      "learning_rate": 6.5000000000000004e-06,
      "loss": 0.0401,
      "step": 8352
    },
    {
      "epoch": 87.0,
      "eval_LOCATION_f1": 0.9072164948453608,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8712871287128713,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03983263671398163,
      "eval_overall_accuracy": 0.9876543209876543,
      "eval_overall_f1": 0.9310344827586207,
      "eval_overall_precision": 0.9197080291970803,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2817,
      "eval_samples_per_second": 603.464,
      "eval_steps_per_second": 10.649,
      "step": 8352
    },
    {
      "epoch": 88.0,
      "grad_norm": 0.568331241607666,
      "learning_rate": 6e-06,
      "loss": 0.041,
      "step": 8448
    },
    {
      "epoch": 88.0,
      "eval_LOCATION_f1": 0.9072164948453608,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8712871287128713,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9161676646706587,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9107142857142857,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.039770666509866714,
      "eval_overall_accuracy": 0.9879286694101509,
      "eval_overall_f1": 0.9334975369458128,
      "eval_overall_precision": 0.9221411192214112,
      "eval_overall_recall": 0.9451371571072319,
      "eval_runtime": 0.2825,
      "eval_samples_per_second": 601.794,
      "eval_steps_per_second": 10.62,
      "step": 8448
    },
    {
      "epoch": 89.0,
      "grad_norm": 1.1911604404449463,
      "learning_rate": 5.500000000000001e-06,
      "loss": 0.0397,
      "step": 8544
    },
    {
      "epoch": 89.0,
      "eval_LOCATION_f1": 0.9015544041450777,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.87,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.039621464908123016,
      "eval_overall_accuracy": 0.9876543209876543,
      "eval_overall_f1": 0.9297163995067818,
      "eval_overall_precision": 0.9195121951219513,
      "eval_overall_recall": 0.940149625935162,
      "eval_runtime": 0.2798,
      "eval_samples_per_second": 607.512,
      "eval_steps_per_second": 10.721,
      "step": 8544
    },
    {
      "epoch": 90.0,
      "grad_norm": 2.4099748134613037,
      "learning_rate": 5e-06,
      "loss": 0.0398,
      "step": 8640
    },
    {
      "epoch": 90.0,
      "eval_LOCATION_f1": 0.9015544041450777,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.87,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9134328358208955,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9053254437869822,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03956400975584984,
      "eval_overall_accuracy": 0.9879286694101509,
      "eval_overall_f1": 0.9310344827586207,
      "eval_overall_precision": 0.9197080291970803,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2819,
      "eval_samples_per_second": 603.1,
      "eval_steps_per_second": 10.643,
      "step": 8640
    },
    {
      "epoch": 91.0,
      "grad_norm": 0.42839816212654114,
      "learning_rate": 4.5e-06,
      "loss": 0.039,
      "step": 8736
    },
    {
      "epoch": 91.0,
      "eval_LOCATION_f1": 0.911917098445596,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.88,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9189189189189191,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9161676646706587,
      "eval_ORGANIZATION_recall": 0.9216867469879518,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03948313742876053,
      "eval_overall_accuracy": 0.9876543209876543,
      "eval_overall_f1": 0.9358024691358026,
      "eval_overall_precision": 0.9266503667481663,
      "eval_overall_recall": 0.9451371571072319,
      "eval_runtime": 0.285,
      "eval_samples_per_second": 596.519,
      "eval_steps_per_second": 10.527,
      "step": 8736
    },
    {
      "epoch": 92.0,
      "grad_norm": 0.5673728585243225,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0385,
      "step": 8832
    },
    {
      "epoch": 92.0,
      "eval_LOCATION_f1": 0.911917098445596,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.88,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03977023437619209,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9321824907521579,
      "eval_overall_precision": 0.9219512195121952,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2816,
      "eval_samples_per_second": 603.659,
      "eval_steps_per_second": 10.653,
      "step": 8832
    },
    {
      "epoch": 93.0,
      "grad_norm": 0.6550965309143066,
      "learning_rate": 3.5000000000000004e-06,
      "loss": 0.0385,
      "step": 8928
    },
    {
      "epoch": 93.0,
      "eval_LOCATION_f1": 0.911917098445596,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.88,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.0398373156785965,
      "eval_overall_accuracy": 0.9873799725651577,
      "eval_overall_f1": 0.9321824907521579,
      "eval_overall_precision": 0.9219512195121952,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2844,
      "eval_samples_per_second": 597.851,
      "eval_steps_per_second": 10.55,
      "step": 8928
    },
    {
      "epoch": 94.0,
      "grad_norm": 0.17871248722076416,
      "learning_rate": 3e-06,
      "loss": 0.0398,
      "step": 9024
    },
    {
      "epoch": 94.0,
      "eval_LOCATION_f1": 0.911917098445596,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.88,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03970322385430336,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9321824907521579,
      "eval_overall_precision": 0.9219512195121952,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2828,
      "eval_samples_per_second": 601.056,
      "eval_steps_per_second": 10.607,
      "step": 9024
    },
    {
      "epoch": 95.0,
      "grad_norm": 0.2635639011859894,
      "learning_rate": 2.5e-06,
      "loss": 0.0382,
      "step": 9120
    },
    {
      "epoch": 95.0,
      "eval_LOCATION_f1": 0.911917098445596,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.88,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03960025683045387,
      "eval_overall_accuracy": 0.9873799725651577,
      "eval_overall_f1": 0.9321824907521579,
      "eval_overall_precision": 0.9219512195121952,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2838,
      "eval_samples_per_second": 599.017,
      "eval_steps_per_second": 10.571,
      "step": 9120
    },
    {
      "epoch": 96.0,
      "grad_norm": 0.5545527935028076,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0408,
      "step": 9216
    },
    {
      "epoch": 96.0,
      "eval_LOCATION_f1": 0.911917098445596,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.88,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.039429329335689545,
      "eval_overall_accuracy": 0.9873799725651577,
      "eval_overall_f1": 0.9321824907521579,
      "eval_overall_precision": 0.9219512195121952,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2843,
      "eval_samples_per_second": 597.868,
      "eval_steps_per_second": 10.551,
      "step": 9216
    },
    {
      "epoch": 97.0,
      "grad_norm": 0.20971649885177612,
      "learning_rate": 1.5e-06,
      "loss": 0.0372,
      "step": 9312
    },
    {
      "epoch": 97.0,
      "eval_LOCATION_f1": 0.911917098445596,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.88,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03945162892341614,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9321824907521579,
      "eval_overall_precision": 0.9219512195121952,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2901,
      "eval_samples_per_second": 586.028,
      "eval_steps_per_second": 10.342,
      "step": 9312
    },
    {
      "epoch": 98.0,
      "grad_norm": 0.9901926517486572,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0392,
      "step": 9408
    },
    {
      "epoch": 98.0,
      "eval_LOCATION_f1": 0.90625,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.8787878787878788,
      "eval_LOCATION_recall": 0.9354838709677419,
      "eval_ORGANIZATION_f1": 0.9074626865671641,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.8994082840236687,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.039484377950429916,
      "eval_overall_accuracy": 0.9871056241426612,
      "eval_overall_f1": 0.9297163995067818,
      "eval_overall_precision": 0.9195121951219513,
      "eval_overall_recall": 0.940149625935162,
      "eval_runtime": 0.2824,
      "eval_samples_per_second": 602.087,
      "eval_steps_per_second": 10.625,
      "step": 9408
    },
    {
      "epoch": 99.0,
      "grad_norm": 0.7696068286895752,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0393,
      "step": 9504
    },
    {
      "epoch": 99.0,
      "eval_LOCATION_f1": 0.911917098445596,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.88,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03950231894850731,
      "eval_overall_accuracy": 0.9873799725651577,
      "eval_overall_f1": 0.9321824907521579,
      "eval_overall_precision": 0.9219512195121952,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2816,
      "eval_samples_per_second": 603.609,
      "eval_steps_per_second": 10.652,
      "step": 9504
    },
    {
      "epoch": 100.0,
      "grad_norm": 0.9257699251174927,
      "learning_rate": 0.0,
      "loss": 0.0399,
      "step": 9600
    },
    {
      "epoch": 100.0,
      "eval_LOCATION_f1": 0.911917098445596,
      "eval_LOCATION_number": 93,
      "eval_LOCATION_precision": 0.88,
      "eval_LOCATION_recall": 0.946236559139785,
      "eval_ORGANIZATION_f1": 0.9101796407185628,
      "eval_ORGANIZATION_number": 166,
      "eval_ORGANIZATION_precision": 0.9047619047619048,
      "eval_ORGANIZATION_recall": 0.9156626506024096,
      "eval_PERSON_f1": 0.971830985915493,
      "eval_PERSON_number": 142,
      "eval_PERSON_precision": 0.971830985915493,
      "eval_PERSON_recall": 0.971830985915493,
      "eval_loss": 0.03953361511230469,
      "eval_overall_accuracy": 0.9873799725651577,
      "eval_overall_f1": 0.9321824907521579,
      "eval_overall_precision": 0.9219512195121952,
      "eval_overall_recall": 0.942643391521197,
      "eval_runtime": 0.2824,
      "eval_samples_per_second": 601.932,
      "eval_steps_per_second": 10.622,
      "step": 9600
    },
    {
      "epoch": 100.0,
      "step": 9600,
      "total_flos": 3784810278471630.0,
      "train_loss": 0.09147172088424364,
      "train_runtime": 505.6149,
      "train_samples_per_second": 302.8,
      "train_steps_per_second": 18.987
    }
  ],
  "logging_steps": 500,
  "max_steps": 9600,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 100,
  "save_steps": 500,
  "total_flos": 3784810278471630.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}