trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 100.0,
  "eval_steps": 500,
  "global_step": 10600,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 1.7324131727218628,
      "learning_rate": 4.9500000000000004e-05,
      "loss": 1.2621,
      "step": 106
    },
    {
      "epoch": 1.0,
      "eval_LOCATION_f1": 0.0,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.0,
      "eval_LOCATION_recall": 0.0,
      "eval_ORGANIZATION_f1": 0.0,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.0,
      "eval_ORGANIZATION_recall": 0.0,
      "eval_PERSON_f1": 0.0,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.0,
      "eval_PERSON_recall": 0.0,
      "eval_QUANTITY_f1": 0.0,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.0,
      "eval_QUANTITY_recall": 0.0,
      "eval_TIME_f1": 0.0,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.0,
      "eval_TIME_recall": 0.0,
      "eval_loss": 0.7093822360038757,
      "eval_overall_accuracy": 0.8448836627470603,
      "eval_overall_f1": 0.0,
      "eval_overall_precision": 0.0,
      "eval_overall_recall": 0.0,
      "eval_runtime": 1.1248,
      "eval_samples_per_second": 166.252,
      "eval_steps_per_second": 2.667,
      "step": 106
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.321110725402832,
      "learning_rate": 4.9e-05,
      "loss": 0.7066,
      "step": 212
    },
    {
      "epoch": 2.0,
      "eval_LOCATION_f1": 0.0,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.0,
      "eval_LOCATION_recall": 0.0,
      "eval_ORGANIZATION_f1": 0.0,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.0,
      "eval_ORGANIZATION_recall": 0.0,
      "eval_PERSON_f1": 0.0,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.0,
      "eval_PERSON_recall": 0.0,
      "eval_QUANTITY_f1": 0.0,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.0,
      "eval_QUANTITY_recall": 0.0,
      "eval_TIME_f1": 0.0,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.0,
      "eval_TIME_recall": 0.0,
      "eval_loss": 0.5970483422279358,
      "eval_overall_accuracy": 0.845634225669252,
      "eval_overall_f1": 0.0,
      "eval_overall_precision": 0.0,
      "eval_overall_recall": 0.0,
      "eval_runtime": 0.9359,
      "eval_samples_per_second": 199.818,
      "eval_steps_per_second": 3.206,
      "step": 212
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.2591755390167236,
      "learning_rate": 4.85e-05,
      "loss": 0.6211,
      "step": 318
    },
    {
      "epoch": 3.0,
      "eval_LOCATION_f1": 0.02702702702702703,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.5,
      "eval_LOCATION_recall": 0.013888888888888888,
      "eval_ORGANIZATION_f1": 0.0,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.0,
      "eval_ORGANIZATION_recall": 0.0,
      "eval_PERSON_f1": 0.06211180124223603,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.2777777777777778,
      "eval_PERSON_recall": 0.03496503496503497,
      "eval_QUANTITY_f1": 0.0,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.0,
      "eval_QUANTITY_recall": 0.0,
      "eval_TIME_f1": 0.0,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.0,
      "eval_TIME_recall": 0.0,
      "eval_loss": 0.5052209496498108,
      "eval_overall_accuracy": 0.8516387290467851,
      "eval_overall_f1": 0.03252032520325203,
      "eval_overall_precision": 0.2,
      "eval_overall_recall": 0.017699115044247787,
      "eval_runtime": 0.9647,
      "eval_samples_per_second": 193.841,
      "eval_steps_per_second": 3.11,
      "step": 318
    },
    {
      "epoch": 4.0,
      "grad_norm": 1.9562969207763672,
      "learning_rate": 4.8e-05,
      "loss": 0.5308,
      "step": 424
    },
    {
      "epoch": 4.0,
      "eval_LOCATION_f1": 0.14457831325301204,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.5454545454545454,
      "eval_LOCATION_recall": 0.08333333333333333,
      "eval_ORGANIZATION_f1": 0.02197802197802198,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.0625,
      "eval_ORGANIZATION_recall": 0.013333333333333334,
      "eval_PERSON_f1": 0.31666666666666665,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.3917525773195876,
      "eval_PERSON_recall": 0.26573426573426573,
      "eval_QUANTITY_f1": 0.0,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.0,
      "eval_QUANTITY_recall": 0.0,
      "eval_TIME_f1": 0.3181818181818182,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.3888888888888889,
      "eval_TIME_recall": 0.2692307692307692,
      "eval_loss": 0.4113342761993408,
      "eval_overall_accuracy": 0.879909932449337,
      "eval_overall_f1": 0.2162162162162162,
      "eval_overall_precision": 0.36619718309859156,
      "eval_overall_recall": 0.15339233038348082,
      "eval_runtime": 0.9301,
      "eval_samples_per_second": 201.053,
      "eval_steps_per_second": 3.225,
      "step": 424
    },
    {
      "epoch": 5.0,
      "grad_norm": 1.885956883430481,
      "learning_rate": 4.75e-05,
      "loss": 0.427,
      "step": 530
    },
    {
      "epoch": 5.0,
      "eval_LOCATION_f1": 0.4424778761061947,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6097560975609756,
      "eval_LOCATION_recall": 0.3472222222222222,
      "eval_ORGANIZATION_f1": 0.27272727272727276,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.42857142857142855,
      "eval_ORGANIZATION_recall": 0.2,
      "eval_PERSON_f1": 0.6179401993355481,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.5886075949367089,
      "eval_PERSON_recall": 0.6503496503496503,
      "eval_QUANTITY_f1": 0.0,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.0,
      "eval_QUANTITY_recall": 0.0,
      "eval_TIME_f1": 0.6382978723404256,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7142857142857143,
      "eval_TIME_recall": 0.5769230769230769,
      "eval_loss": 0.3248092830181122,
      "eval_overall_accuracy": 0.9101826369777333,
      "eval_overall_f1": 0.49006622516556286,
      "eval_overall_precision": 0.5584905660377358,
      "eval_overall_recall": 0.4365781710914454,
      "eval_runtime": 0.8911,
      "eval_samples_per_second": 209.848,
      "eval_steps_per_second": 3.367,
      "step": 530
    },
    {
      "epoch": 6.0,
      "grad_norm": 1.2173807621002197,
      "learning_rate": 4.7e-05,
      "loss": 0.3429,
      "step": 636
    },
    {
      "epoch": 6.0,
      "eval_LOCATION_f1": 0.593103448275862,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.589041095890411,
      "eval_LOCATION_recall": 0.5972222222222222,
      "eval_ORGANIZATION_f1": 0.4729729729729729,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.4794520547945205,
      "eval_ORGANIZATION_recall": 0.4666666666666667,
      "eval_PERSON_f1": 0.7538940809968847,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.6797752808988764,
      "eval_PERSON_recall": 0.8461538461538461,
      "eval_QUANTITY_f1": 0.15999999999999998,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.14814814814814814,
      "eval_QUANTITY_recall": 0.17391304347826086,
      "eval_TIME_f1": 0.6666666666666666,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6428571428571429,
      "eval_TIME_recall": 0.6923076923076923,
      "eval_loss": 0.2732473909854889,
      "eval_overall_accuracy": 0.9274455841881412,
      "eval_overall_f1": 0.6155988857938719,
      "eval_overall_precision": 0.58311345646438,
      "eval_overall_recall": 0.6519174041297935,
      "eval_runtime": 0.887,
      "eval_samples_per_second": 210.819,
      "eval_steps_per_second": 3.382,
      "step": 636
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.8688632845878601,
      "learning_rate": 4.6500000000000005e-05,
      "loss": 0.2873,
      "step": 742
    },
    {
      "epoch": 7.0,
      "eval_LOCATION_f1": 0.6623376623376623,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6219512195121951,
      "eval_LOCATION_recall": 0.7083333333333334,
      "eval_ORGANIZATION_f1": 0.4968152866242038,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.47560975609756095,
      "eval_ORGANIZATION_recall": 0.52,
      "eval_PERSON_f1": 0.80625,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7288135593220338,
      "eval_PERSON_recall": 0.9020979020979021,
      "eval_QUANTITY_f1": 0.2962962962962963,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.25806451612903225,
      "eval_QUANTITY_recall": 0.34782608695652173,
      "eval_TIME_f1": 0.7272727272727274,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6896551724137931,
      "eval_TIME_recall": 0.7692307692307693,
      "eval_loss": 0.23371461033821106,
      "eval_overall_accuracy": 0.9329497122842132,
      "eval_overall_f1": 0.6675675675675676,
      "eval_overall_precision": 0.6159600997506235,
      "eval_overall_recall": 0.7286135693215339,
      "eval_runtime": 0.9045,
      "eval_samples_per_second": 206.749,
      "eval_steps_per_second": 3.317,
      "step": 742
    },
    {
      "epoch": 8.0,
      "grad_norm": 1.5551362037658691,
      "learning_rate": 4.600000000000001e-05,
      "loss": 0.249,
      "step": 848
    },
    {
      "epoch": 8.0,
      "eval_LOCATION_f1": 0.732919254658385,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6629213483146067,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.5813953488372093,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5154639175257731,
      "eval_ORGANIZATION_recall": 0.6666666666666666,
      "eval_PERSON_f1": 0.832807570977918,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7586206896551724,
      "eval_PERSON_recall": 0.9230769230769231,
      "eval_QUANTITY_f1": 0.2962962962962963,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.25806451612903225,
      "eval_QUANTITY_recall": 0.34782608695652173,
      "eval_TIME_f1": 0.7272727272727274,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6896551724137931,
      "eval_TIME_recall": 0.7692307692307693,
      "eval_loss": 0.2129368931055069,
      "eval_overall_accuracy": 0.9367025268951714,
      "eval_overall_f1": 0.7088274044795784,
      "eval_overall_precision": 0.6404761904761904,
      "eval_overall_recall": 0.7935103244837758,
      "eval_runtime": 0.9166,
      "eval_samples_per_second": 204.019,
      "eval_steps_per_second": 3.273,
      "step": 848
    },
    {
      "epoch": 9.0,
      "grad_norm": 1.3527239561080933,
      "learning_rate": 4.55e-05,
      "loss": 0.2259,
      "step": 954
    },
    {
      "epoch": 9.0,
      "eval_LOCATION_f1": 0.7625000000000001,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6931818181818182,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.6046511627906977,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5360824742268041,
      "eval_ORGANIZATION_recall": 0.6933333333333334,
      "eval_PERSON_f1": 0.8360128617363344,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7738095238095238,
      "eval_PERSON_recall": 0.9090909090909091,
      "eval_QUANTITY_f1": 0.3103448275862069,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.2571428571428571,
      "eval_QUANTITY_recall": 0.391304347826087,
      "eval_TIME_f1": 0.7272727272727274,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6896551724137931,
      "eval_TIME_recall": 0.7692307692307693,
      "eval_loss": 0.19730736315250397,
      "eval_overall_accuracy": 0.9384538403802852,
      "eval_overall_f1": 0.7195767195767195,
      "eval_overall_precision": 0.6522781774580336,
      "eval_overall_recall": 0.8023598820058997,
      "eval_runtime": 0.9257,
      "eval_samples_per_second": 202.019,
      "eval_steps_per_second": 3.241,
      "step": 954
    },
    {
      "epoch": 10.0,
      "grad_norm": 2.4482581615448,
      "learning_rate": 4.5e-05,
      "loss": 0.2088,
      "step": 1060
    },
    {
      "epoch": 10.0,
      "eval_LOCATION_f1": 0.7654320987654321,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6888888888888889,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.5847953216374269,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5208333333333334,
      "eval_ORGANIZATION_recall": 0.6666666666666666,
      "eval_PERSON_f1": 0.840764331210191,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7719298245614035,
      "eval_PERSON_recall": 0.9230769230769231,
      "eval_QUANTITY_f1": 0.4137931034482759,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.34285714285714286,
      "eval_QUANTITY_recall": 0.5217391304347826,
      "eval_TIME_f1": 0.6666666666666667,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6129032258064516,
      "eval_TIME_recall": 0.7307692307692307,
      "eval_loss": 0.18520984053611755,
      "eval_overall_accuracy": 0.9392044033024769,
      "eval_overall_f1": 0.7217847769028872,
      "eval_overall_precision": 0.6501182033096927,
      "eval_overall_recall": 0.8112094395280236,
      "eval_runtime": 0.9227,
      "eval_samples_per_second": 202.666,
      "eval_steps_per_second": 3.251,
      "step": 1060
    },
    {
      "epoch": 11.0,
      "grad_norm": 1.6825419664382935,
      "learning_rate": 4.4500000000000004e-05,
      "loss": 0.196,
      "step": 1166
    },
    {
      "epoch": 11.0,
      "eval_LOCATION_f1": 0.7974683544303797,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7325581395348837,
      "eval_LOCATION_recall": 0.875,
      "eval_ORGANIZATION_f1": 0.5868263473053893,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.532608695652174,
      "eval_ORGANIZATION_recall": 0.6533333333333333,
      "eval_PERSON_f1": 0.8488745980707394,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7857142857142857,
      "eval_PERSON_recall": 0.9230769230769231,
      "eval_QUANTITY_f1": 0.3928571428571428,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.3333333333333333,
      "eval_QUANTITY_recall": 0.4782608695652174,
      "eval_TIME_f1": 0.7142857142857142,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6666666666666666,
      "eval_TIME_recall": 0.7692307692307693,
      "eval_loss": 0.1723126471042633,
      "eval_overall_accuracy": 0.9434575931948962,
      "eval_overall_f1": 0.7352941176470589,
      "eval_overall_precision": 0.6723716381418093,
      "eval_overall_recall": 0.8112094395280236,
      "eval_runtime": 0.9093,
      "eval_samples_per_second": 205.658,
      "eval_steps_per_second": 3.299,
      "step": 1166
    },
    {
      "epoch": 12.0,
      "grad_norm": 2.3418383598327637,
      "learning_rate": 4.4000000000000006e-05,
      "loss": 0.1884,
      "step": 1272
    },
    {
      "epoch": 12.0,
      "eval_LOCATION_f1": 0.7973856209150327,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7530864197530864,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.6473988439306357,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5714285714285714,
      "eval_ORGANIZATION_recall": 0.7466666666666667,
      "eval_PERSON_f1": 0.8461538461538461,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7810650887573964,
      "eval_PERSON_recall": 0.9230769230769231,
      "eval_QUANTITY_f1": 0.4814814814814815,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.41935483870967744,
      "eval_QUANTITY_recall": 0.5652173913043478,
      "eval_TIME_f1": 0.7,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6176470588235294,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1679195761680603,
      "eval_overall_accuracy": 0.9444583437578183,
      "eval_overall_f1": 0.7526595744680852,
      "eval_overall_precision": 0.6852300242130751,
      "eval_overall_recall": 0.8348082595870207,
      "eval_runtime": 0.9102,
      "eval_samples_per_second": 205.442,
      "eval_steps_per_second": 3.296,
      "step": 1272
    },
    {
      "epoch": 13.0,
      "grad_norm": 1.0822410583496094,
      "learning_rate": 4.35e-05,
      "loss": 0.1791,
      "step": 1378
    },
    {
      "epoch": 13.0,
      "eval_LOCATION_f1": 0.781456953642384,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7468354430379747,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.6744186046511628,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5979381443298969,
      "eval_ORGANIZATION_recall": 0.7733333333333333,
      "eval_PERSON_f1": 0.8488745980707394,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7857142857142857,
      "eval_PERSON_recall": 0.9230769230769231,
      "eval_QUANTITY_f1": 0.5185185185185185,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.45161290322580644,
      "eval_QUANTITY_recall": 0.6086956521739131,
      "eval_TIME_f1": 0.7118644067796609,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6363636363636364,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.16172969341278076,
      "eval_overall_accuracy": 0.9467100325243933,
      "eval_overall_f1": 0.7603748326639893,
      "eval_overall_precision": 0.696078431372549,
      "eval_overall_recall": 0.8377581120943953,
      "eval_runtime": 0.9055,
      "eval_samples_per_second": 206.514,
      "eval_steps_per_second": 3.313,
      "step": 1378
    },
    {
      "epoch": 14.0,
      "grad_norm": 1.5624088048934937,
      "learning_rate": 4.3e-05,
      "loss": 0.1713,
      "step": 1484
    },
    {
      "epoch": 14.0,
      "eval_LOCATION_f1": 0.7837837837837838,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7631578947368421,
      "eval_LOCATION_recall": 0.8055555555555556,
      "eval_ORGANIZATION_f1": 0.6704545454545454,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5841584158415841,
      "eval_ORGANIZATION_recall": 0.7866666666666666,
      "eval_PERSON_f1": 0.8636363636363636,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.806060606060606,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.5098039215686274,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.4642857142857143,
      "eval_QUANTITY_recall": 0.5652173913043478,
      "eval_TIME_f1": 0.7857142857142856,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7333333333333333,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.15461112558841705,
      "eval_overall_accuracy": 0.9497122842131599,
      "eval_overall_f1": 0.7713125845737482,
      "eval_overall_precision": 0.7125,
      "eval_overall_recall": 0.8407079646017699,
      "eval_runtime": 0.9063,
      "eval_samples_per_second": 206.328,
      "eval_steps_per_second": 3.31,
      "step": 1484
    },
    {
      "epoch": 15.0,
      "grad_norm": 0.8365355730056763,
      "learning_rate": 4.25e-05,
      "loss": 0.1674,
      "step": 1590
    },
    {
      "epoch": 15.0,
      "eval_LOCATION_f1": 0.7843137254901961,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7407407407407407,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.6704545454545454,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5841584158415841,
      "eval_ORGANIZATION_recall": 0.7866666666666666,
      "eval_PERSON_f1": 0.8543689320388351,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7951807228915663,
      "eval_PERSON_recall": 0.9230769230769231,
      "eval_QUANTITY_f1": 0.49056603773584906,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.43333333333333335,
      "eval_QUANTITY_recall": 0.5652173913043478,
      "eval_TIME_f1": 0.7241379310344829,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.65625,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1564231663942337,
      "eval_overall_accuracy": 0.9494620965724293,
      "eval_overall_f1": 0.761014686248331,
      "eval_overall_precision": 0.6951219512195121,
      "eval_overall_recall": 0.8407079646017699,
      "eval_runtime": 0.9118,
      "eval_samples_per_second": 205.09,
      "eval_steps_per_second": 3.29,
      "step": 1590
    },
    {
      "epoch": 16.0,
      "grad_norm": 1.1776515245437622,
      "learning_rate": 4.2e-05,
      "loss": 0.1623,
      "step": 1696
    },
    {
      "epoch": 16.0,
      "eval_LOCATION_f1": 0.7843137254901961,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7407407407407407,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.6818181818181819,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.594059405940594,
      "eval_ORGANIZATION_recall": 0.8,
      "eval_PERSON_f1": 0.8571428571428571,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8,
      "eval_PERSON_recall": 0.9230769230769231,
      "eval_QUANTITY_f1": 0.6296296296296297,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5483870967741935,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7719298245614036,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7096774193548387,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.15418055653572083,
      "eval_overall_accuracy": 0.9509632224168126,
      "eval_overall_f1": 0.7780748663101604,
      "eval_overall_precision": 0.7114914425427873,
      "eval_overall_recall": 0.8584070796460177,
      "eval_runtime": 0.9082,
      "eval_samples_per_second": 205.891,
      "eval_steps_per_second": 3.303,
      "step": 1696
    },
    {
      "epoch": 17.0,
      "grad_norm": 0.6350365877151489,
      "learning_rate": 4.15e-05,
      "loss": 0.158,
      "step": 1802
    },
    {
      "epoch": 17.0,
      "eval_LOCATION_f1": 0.751592356687898,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6941176470588235,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.6666666666666666,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5784313725490197,
      "eval_ORGANIZATION_recall": 0.7866666666666666,
      "eval_PERSON_f1": 0.8636363636363636,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.806060606060606,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7719298245614036,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7096774193548387,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.15079031884670258,
      "eval_overall_accuracy": 0.950212659494621,
      "eval_overall_f1": 0.7708609271523179,
      "eval_overall_precision": 0.6995192307692307,
      "eval_overall_recall": 0.8584070796460177,
      "eval_runtime": 0.9247,
      "eval_samples_per_second": 202.221,
      "eval_steps_per_second": 3.244,
      "step": 1802
    },
    {
      "epoch": 18.0,
      "grad_norm": 0.8113768100738525,
      "learning_rate": 4.1e-05,
      "loss": 0.1518,
      "step": 1908
    },
    {
      "epoch": 18.0,
      "eval_LOCATION_f1": 0.759493670886076,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6976744186046512,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.6666666666666666,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5784313725490197,
      "eval_ORGANIZATION_recall": 0.7866666666666666,
      "eval_PERSON_f1": 0.8580645161290323,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7964071856287425,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7719298245614036,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7096774193548387,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1505645364522934,
      "eval_overall_accuracy": 0.9497122842131599,
      "eval_overall_f1": 0.770448548812665,
      "eval_overall_precision": 0.6968973747016707,
      "eval_overall_recall": 0.8613569321533924,
      "eval_runtime": 0.9072,
      "eval_samples_per_second": 206.139,
      "eval_steps_per_second": 3.307,
      "step": 1908
    },
    {
      "epoch": 19.0,
      "grad_norm": 1.1493066549301147,
      "learning_rate": 4.05e-05,
      "loss": 0.1508,
      "step": 2014
    },
    {
      "epoch": 19.0,
      "eval_LOCATION_f1": 0.7763157894736842,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7375,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.6820809248554913,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6020408163265306,
      "eval_ORGANIZATION_recall": 0.7866666666666666,
      "eval_PERSON_f1": 0.8721311475409836,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8209876543209876,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.14393331110477448,
      "eval_overall_accuracy": 0.9532149111833875,
      "eval_overall_f1": 0.7881241565452092,
      "eval_overall_precision": 0.7263681592039801,
      "eval_overall_recall": 0.8613569321533924,
      "eval_runtime": 0.9095,
      "eval_samples_per_second": 205.605,
      "eval_steps_per_second": 3.298,
      "step": 2014
    },
    {
      "epoch": 20.0,
      "grad_norm": 1.1086537837982178,
      "learning_rate": 4e-05,
      "loss": 0.1435,
      "step": 2120
    },
    {
      "epoch": 20.0,
      "eval_LOCATION_f1": 0.7848101265822784,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7209302325581395,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.6666666666666666,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5585585585585585,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8627450980392156,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8098159509202454,
      "eval_PERSON_recall": 0.9230769230769231,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7719298245614036,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7096774193548387,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.15614458918571472,
      "eval_overall_accuracy": 0.950212659494621,
      "eval_overall_f1": 0.7769028871391076,
      "eval_overall_precision": 0.6997635933806147,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9172,
      "eval_samples_per_second": 203.874,
      "eval_steps_per_second": 3.271,
      "step": 2120
    },
    {
      "epoch": 21.0,
      "grad_norm": 0.9740850925445557,
      "learning_rate": 3.9500000000000005e-05,
      "loss": 0.1456,
      "step": 2226
    },
    {
      "epoch": 21.0,
      "eval_LOCATION_f1": 0.7662337662337662,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.6705202312138728,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5918367346938775,
      "eval_ORGANIZATION_recall": 0.7733333333333333,
      "eval_PERSON_f1": 0.8721311475409836,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8209876543209876,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.6296296296296297,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5483870967741935,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8214285714285715,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7666666666666667,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.14360801875591278,
      "eval_overall_accuracy": 0.9537152864648486,
      "eval_overall_f1": 0.7816711590296497,
      "eval_overall_precision": 0.7196029776674938,
      "eval_overall_recall": 0.855457227138643,
      "eval_runtime": 0.9095,
      "eval_samples_per_second": 205.608,
      "eval_steps_per_second": 3.299,
      "step": 2226
    },
    {
      "epoch": 22.0,
      "grad_norm": 0.8203905820846558,
      "learning_rate": 3.9000000000000006e-05,
      "loss": 0.1384,
      "step": 2332
    },
    {
      "epoch": 22.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7126436781609197,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6262626262626263,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8758169934640523,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8220858895705522,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8214285714285715,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7666666666666667,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.14492544531822205,
      "eval_overall_accuracy": 0.9542156617463098,
      "eval_overall_f1": 0.8016085790884719,
      "eval_overall_precision": 0.7346437346437347,
      "eval_overall_recall": 0.8820058997050148,
      "eval_runtime": 0.9167,
      "eval_samples_per_second": 203.986,
      "eval_steps_per_second": 3.273,
      "step": 2332
    },
    {
      "epoch": 23.0,
      "grad_norm": 5.496976375579834,
      "learning_rate": 3.85e-05,
      "loss": 0.1384,
      "step": 2438
    },
    {
      "epoch": 23.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7011494252873564,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6161616161616161,
      "eval_ORGANIZATION_recall": 0.8133333333333334,
      "eval_PERSON_f1": 0.8786885245901639,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8271604938271605,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6666666666666667,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5806451612903226,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.14144913852214813,
      "eval_overall_accuracy": 0.9562171628721541,
      "eval_overall_f1": 0.8005390835579514,
      "eval_overall_precision": 0.7369727047146402,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9123,
      "eval_samples_per_second": 204.982,
      "eval_steps_per_second": 3.288,
      "step": 2438
    },
    {
      "epoch": 24.0,
      "grad_norm": 1.060779333114624,
      "learning_rate": 3.8e-05,
      "loss": 0.1369,
      "step": 2544
    },
    {
      "epoch": 24.0,
      "eval_LOCATION_f1": 0.7898089171974522,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7294117647058823,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.6775956284153005,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5740740740740741,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8627450980392156,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8098159509202454,
      "eval_PERSON_recall": 0.9230769230769231,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1503037065267563,
      "eval_overall_accuracy": 0.9509632224168126,
      "eval_overall_f1": 0.7820343461030383,
      "eval_overall_precision": 0.7081339712918661,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9133,
      "eval_samples_per_second": 204.754,
      "eval_steps_per_second": 3.285,
      "step": 2544
    },
    {
      "epoch": 25.0,
      "grad_norm": 0.9608832001686096,
      "learning_rate": 3.7500000000000003e-05,
      "loss": 0.136,
      "step": 2650
    },
    {
      "epoch": 25.0,
      "eval_LOCATION_f1": 0.7870967741935484,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7349397590361446,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7058823529411765,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.631578947368421,
      "eval_ORGANIZATION_recall": 0.8,
      "eval_PERSON_f1": 0.8758169934640523,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8220858895705522,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1423281729221344,
      "eval_overall_accuracy": 0.955716787590693,
      "eval_overall_f1": 0.7951482479784365,
      "eval_overall_precision": 0.7320099255583127,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.9155,
      "eval_samples_per_second": 204.256,
      "eval_steps_per_second": 3.277,
      "step": 2650
    },
    {
      "epoch": 26.0,
      "grad_norm": 1.9149385690689087,
      "learning_rate": 3.7e-05,
      "loss": 0.1322,
      "step": 2756
    },
    {
      "epoch": 26.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7209302325581395,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6391752577319587,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8758169934640523,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8220858895705522,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7857142857142856,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7333333333333333,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.14195844531059265,
      "eval_overall_accuracy": 0.9554665999499625,
      "eval_overall_f1": 0.7994616419919246,
      "eval_overall_precision": 0.7351485148514851,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9182,
      "eval_samples_per_second": 203.651,
      "eval_steps_per_second": 3.267,
      "step": 2756
    },
    {
      "epoch": 27.0,
      "grad_norm": 4.680337905883789,
      "learning_rate": 3.65e-05,
      "loss": 0.1323,
      "step": 2862
    },
    {
      "epoch": 27.0,
      "eval_LOCATION_f1": 0.7948717948717949,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7380952380952381,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7305389221556886,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6630434782608695,
      "eval_ORGANIZATION_recall": 0.8133333333333334,
      "eval_PERSON_f1": 0.8910891089108911,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84375,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.14178352057933807,
      "eval_overall_accuracy": 0.9569677257943457,
      "eval_overall_f1": 0.8059701492537314,
      "eval_overall_precision": 0.7462311557788944,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9164,
      "eval_samples_per_second": 204.054,
      "eval_steps_per_second": 3.274,
      "step": 2862
    },
    {
      "epoch": 28.0,
      "grad_norm": 1.5706373453140259,
      "learning_rate": 3.6e-05,
      "loss": 0.1306,
      "step": 2968
    },
    {
      "epoch": 28.0,
      "eval_LOCATION_f1": 0.7682119205298013,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7341772151898734,
      "eval_LOCATION_recall": 0.8055555555555556,
      "eval_ORGANIZATION_f1": 0.7151515151515152,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6555555555555556,
      "eval_ORGANIZATION_recall": 0.7866666666666666,
      "eval_PERSON_f1": 0.8750000000000001,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8260869565217391,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.6792452830188679,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.6,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8518518518518519,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.8214285714285714,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.13230037689208984,
      "eval_overall_accuracy": 0.9584688516387291,
      "eval_overall_f1": 0.8005502063273728,
      "eval_overall_precision": 0.75,
      "eval_overall_recall": 0.8584070796460177,
      "eval_runtime": 0.9167,
      "eval_samples_per_second": 203.989,
      "eval_steps_per_second": 3.273,
      "step": 2968
    },
    {
      "epoch": 29.0,
      "grad_norm": 1.6984310150146484,
      "learning_rate": 3.55e-05,
      "loss": 0.1282,
      "step": 3074
    },
    {
      "epoch": 29.0,
      "eval_LOCATION_f1": 0.8026315789473685,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7625,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7176470588235295,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6421052631578947,
      "eval_ORGANIZATION_recall": 0.8133333333333334,
      "eval_PERSON_f1": 0.8963210702341137,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8589743589743589,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6792452830188679,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.6,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8518518518518519,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.8214285714285714,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.13585858047008514,
      "eval_overall_accuracy": 0.9584688516387291,
      "eval_overall_f1": 0.8159340659340658,
      "eval_overall_precision": 0.7634961439588689,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9131,
      "eval_samples_per_second": 204.804,
      "eval_steps_per_second": 3.286,
      "step": 3074
    },
    {
      "epoch": 30.0,
      "grad_norm": 0.8933135271072388,
      "learning_rate": 3.5e-05,
      "loss": 0.1235,
      "step": 3180
    },
    {
      "epoch": 30.0,
      "eval_LOCATION_f1": 0.7973856209150327,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7530864197530864,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7159090909090909,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6237623762376238,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13996899127960205,
      "eval_overall_accuracy": 0.9559669752314236,
      "eval_overall_f1": 0.8054054054054054,
      "eval_overall_precision": 0.743142144638404,
      "eval_overall_recall": 0.8790560471976401,
      "eval_runtime": 0.9167,
      "eval_samples_per_second": 203.988,
      "eval_steps_per_second": 3.273,
      "step": 3180
    },
    {
      "epoch": 31.0,
      "grad_norm": 0.8632007837295532,
      "learning_rate": 3.45e-05,
      "loss": 0.124,
      "step": 3286
    },
    {
      "epoch": 31.0,
      "eval_LOCATION_f1": 0.7741935483870969,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7228915662650602,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7176470588235295,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6421052631578947,
      "eval_ORGANIZATION_recall": 0.8133333333333334,
      "eval_PERSON_f1": 0.8874172185430464,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8427672955974843,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.13782481849193573,
      "eval_overall_accuracy": 0.9552164123092319,
      "eval_overall_f1": 0.8021680216802167,
      "eval_overall_precision": 0.7418546365914787,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9178,
      "eval_samples_per_second": 203.741,
      "eval_steps_per_second": 3.269,
      "step": 3286
    },
    {
      "epoch": 32.0,
      "grad_norm": 1.97568941116333,
      "learning_rate": 3.4000000000000007e-05,
      "loss": 0.1222,
      "step": 3392
    },
    {
      "epoch": 32.0,
      "eval_LOCATION_f1": 0.7894736842105262,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.75,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7361963190184048,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6818181818181818,
      "eval_ORGANIZATION_recall": 0.8,
      "eval_PERSON_f1": 0.8896321070234114,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8525641025641025,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.6792452830188679,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.6,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.13005581498146057,
      "eval_overall_accuracy": 0.9597197898423818,
      "eval_overall_f1": 0.8144044321329639,
      "eval_overall_precision": 0.7676240208877284,
      "eval_overall_recall": 0.8672566371681416,
      "eval_runtime": 0.9102,
      "eval_samples_per_second": 205.446,
      "eval_steps_per_second": 3.296,
      "step": 3392
    },
    {
      "epoch": 33.0,
      "grad_norm": 1.2445570230484009,
      "learning_rate": 3.35e-05,
      "loss": 0.1243,
      "step": 3498
    },
    {
      "epoch": 33.0,
      "eval_LOCATION_f1": 0.7870967741935484,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7349397590361446,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7380952380952381,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6666666666666666,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8866666666666668,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8471337579617835,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13504153490066528,
      "eval_overall_accuracy": 0.957968476357268,
      "eval_overall_f1": 0.8076398362892225,
      "eval_overall_precision": 0.751269035532995,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9062,
      "eval_samples_per_second": 206.352,
      "eval_steps_per_second": 3.31,
      "step": 3498
    },
    {
      "epoch": 34.0,
      "grad_norm": 0.8017980456352234,
      "learning_rate": 3.3e-05,
      "loss": 0.1198,
      "step": 3604
    },
    {
      "epoch": 34.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7209302325581395,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6391752577319587,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8852459016393444,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8333333333333334,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.75,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1404941827058792,
      "eval_overall_accuracy": 0.9567175381536153,
      "eval_overall_f1": 0.797843665768194,
      "eval_overall_precision": 0.7344913151364765,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9094,
      "eval_samples_per_second": 205.619,
      "eval_steps_per_second": 3.299,
      "step": 3604
    },
    {
      "epoch": 35.0,
      "grad_norm": 1.849617600440979,
      "learning_rate": 3.2500000000000004e-05,
      "loss": 0.1186,
      "step": 3710
    },
    {
      "epoch": 35.0,
      "eval_LOCATION_f1": 0.7973856209150327,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7530864197530864,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.72,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.63,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1358003318309784,
      "eval_overall_accuracy": 0.9567175381536153,
      "eval_overall_f1": 0.803788903924222,
      "eval_overall_precision": 0.7425,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9073,
      "eval_samples_per_second": 206.098,
      "eval_steps_per_second": 3.306,
      "step": 3710
    },
    {
      "epoch": 36.0,
      "grad_norm": 1.2370284795761108,
      "learning_rate": 3.2000000000000005e-05,
      "loss": 0.1177,
      "step": 3816
    },
    {
      "epoch": 36.0,
      "eval_LOCATION_f1": 0.7763157894736842,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7375,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7272727272727272,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6666666666666666,
      "eval_ORGANIZATION_recall": 0.8,
      "eval_PERSON_f1": 0.8844884488448846,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8375,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.1319529414176941,
      "eval_overall_accuracy": 0.9584688516387291,
      "eval_overall_f1": 0.8027397260273973,
      "eval_overall_precision": 0.7493606138107417,
      "eval_overall_recall": 0.8643067846607669,
      "eval_runtime": 0.9101,
      "eval_samples_per_second": 205.474,
      "eval_steps_per_second": 3.296,
      "step": 3816
    },
    {
      "epoch": 37.0,
      "grad_norm": 2.0164999961853027,
      "learning_rate": 3.15e-05,
      "loss": 0.1183,
      "step": 3922
    },
    {
      "epoch": 37.0,
      "eval_LOCATION_f1": 0.782051282051282,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7261904761904762,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.757396449704142,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6808510638297872,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8910891089108911,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84375,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.13670241832733154,
      "eval_overall_accuracy": 0.9572179134350763,
      "eval_overall_f1": 0.8119079837618404,
      "eval_overall_precision": 0.75,
      "eval_overall_recall": 0.8849557522123894,
      "eval_runtime": 0.9293,
      "eval_samples_per_second": 201.235,
      "eval_steps_per_second": 3.228,
      "step": 3922
    },
    {
      "epoch": 38.0,
      "grad_norm": 1.5912907123565674,
      "learning_rate": 3.1e-05,
      "loss": 0.1145,
      "step": 4028
    },
    {
      "epoch": 38.0,
      "eval_LOCATION_f1": 0.7763157894736842,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7375,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.75,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6774193548387096,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8794788273615636,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.823170731707317,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.5964912280701754,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13946709036827087,
      "eval_overall_accuracy": 0.9562171628721541,
      "eval_overall_f1": 0.7983761840324762,
      "eval_overall_precision": 0.7375,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.918,
      "eval_samples_per_second": 203.705,
      "eval_steps_per_second": 3.268,
      "step": 4028
    },
    {
      "epoch": 39.0,
      "grad_norm": 0.7153412103652954,
      "learning_rate": 3.05e-05,
      "loss": 0.1138,
      "step": 4134
    },
    {
      "epoch": 39.0,
      "eval_LOCATION_f1": 0.8133333333333332,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.782051282051282,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7251461988304094,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6458333333333334,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8963210702341137,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8589743589743589,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.1324165016412735,
      "eval_overall_accuracy": 0.957968476357268,
      "eval_overall_f1": 0.810958904109589,
      "eval_overall_precision": 0.7570332480818415,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9158,
      "eval_samples_per_second": 204.202,
      "eval_steps_per_second": 3.276,
      "step": 4134
    },
    {
      "epoch": 40.0,
      "grad_norm": 1.3694576025009155,
      "learning_rate": 3e-05,
      "loss": 0.1131,
      "step": 4240
    },
    {
      "epoch": 40.0,
      "eval_LOCATION_f1": 0.781456953642384,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7468354430379747,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7469879518072289,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6813186813186813,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8940397350993377,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8490566037735849,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13166731595993042,
      "eval_overall_accuracy": 0.9584688516387291,
      "eval_overall_f1": 0.8049450549450551,
      "eval_overall_precision": 0.7532133676092545,
      "eval_overall_recall": 0.8643067846607669,
      "eval_runtime": 0.9126,
      "eval_samples_per_second": 204.913,
      "eval_steps_per_second": 3.287,
      "step": 4240
    },
    {
      "epoch": 41.0,
      "grad_norm": 1.4050893783569336,
      "learning_rate": 2.95e-05,
      "loss": 0.1123,
      "step": 4346
    },
    {
      "epoch": 41.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7529411764705883,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6736842105263158,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8729641693811075,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8170731707317073,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13514545559883118,
      "eval_overall_accuracy": 0.9569677257943457,
      "eval_overall_f1": 0.7989203778677463,
      "eval_overall_precision": 0.736318407960199,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9113,
      "eval_samples_per_second": 205.195,
      "eval_steps_per_second": 3.292,
      "step": 4346
    },
    {
      "epoch": 42.0,
      "grad_norm": 1.5570241212844849,
      "learning_rate": 2.9e-05,
      "loss": 0.1123,
      "step": 4452
    },
    {
      "epoch": 42.0,
      "eval_LOCATION_f1": 0.7973856209150327,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7530864197530864,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7425149700598803,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6739130434782609,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8852459016393444,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8333333333333334,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.5964912280701754,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13604854047298431,
      "eval_overall_accuracy": 0.9572179134350763,
      "eval_overall_f1": 0.8059701492537314,
      "eval_overall_precision": 0.7462311557788944,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9129,
      "eval_samples_per_second": 204.836,
      "eval_steps_per_second": 3.286,
      "step": 4452
    },
    {
      "epoch": 43.0,
      "grad_norm": 0.9890122413635254,
      "learning_rate": 2.8499999999999998e-05,
      "loss": 0.1106,
      "step": 4558
    },
    {
      "epoch": 43.0,
      "eval_LOCATION_f1": 0.7763157894736842,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7375,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7425149700598803,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6739130434782609,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8322981366459627,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13483330607414246,
      "eval_overall_accuracy": 0.9572179134350763,
      "eval_overall_f1": 0.8021828103683494,
      "eval_overall_precision": 0.7461928934010152,
      "eval_overall_recall": 0.8672566371681416,
      "eval_runtime": 0.9143,
      "eval_samples_per_second": 204.536,
      "eval_steps_per_second": 3.281,
      "step": 4558
    },
    {
      "epoch": 44.0,
      "grad_norm": 1.2907875776290894,
      "learning_rate": 2.8000000000000003e-05,
      "loss": 0.1104,
      "step": 4664
    },
    {
      "epoch": 44.0,
      "eval_LOCATION_f1": 0.7894736842105262,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.75,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7484662576687117,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6931818181818182,
      "eval_ORGANIZATION_recall": 0.8133333333333334,
      "eval_PERSON_f1": 0.881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8322981366459627,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6153846153846154,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5517241379310345,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.13180701434612274,
      "eval_overall_accuracy": 0.9589692269201902,
      "eval_overall_f1": 0.8099173553719009,
      "eval_overall_precision": 0.7596899224806202,
      "eval_overall_recall": 0.8672566371681416,
      "eval_runtime": 0.9205,
      "eval_samples_per_second": 203.159,
      "eval_steps_per_second": 3.259,
      "step": 4664
    },
    {
      "epoch": 45.0,
      "grad_norm": 1.4953218698501587,
      "learning_rate": 2.7500000000000004e-05,
      "loss": 0.1084,
      "step": 4770
    },
    {
      "epoch": 45.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7294117647058824,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6526315789473685,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8947368421052632,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84472049689441,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13945060968399048,
      "eval_overall_accuracy": 0.9554665999499625,
      "eval_overall_f1": 0.8010825439783491,
      "eval_overall_precision": 0.74,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9155,
      "eval_samples_per_second": 204.271,
      "eval_steps_per_second": 3.277,
      "step": 4770
    },
    {
      "epoch": 46.0,
      "grad_norm": 0.8582813143730164,
      "learning_rate": 2.7000000000000002e-05,
      "loss": 0.1084,
      "step": 4876
    },
    {
      "epoch": 46.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7619047619047621,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6881720430107527,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8385093167701864,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6415094339622641,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5666666666666667,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1349920630455017,
      "eval_overall_accuracy": 0.957968476357268,
      "eval_overall_f1": 0.8147138964577656,
      "eval_overall_precision": 0.7569620253164557,
      "eval_overall_recall": 0.8820058997050148,
      "eval_runtime": 0.9125,
      "eval_samples_per_second": 204.932,
      "eval_steps_per_second": 3.288,
      "step": 4876
    },
    {
      "epoch": 47.0,
      "grad_norm": 0.8841750621795654,
      "learning_rate": 2.6500000000000004e-05,
      "loss": 0.109,
      "step": 4982
    },
    {
      "epoch": 47.0,
      "eval_LOCATION_f1": 0.7741935483870969,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7228915662650602,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.736842105263158,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.65625,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8910891089108911,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84375,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13927918672561646,
      "eval_overall_accuracy": 0.955716787590693,
      "eval_overall_f1": 0.7999999999999999,
      "eval_overall_precision": 0.7381546134663342,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.904,
      "eval_samples_per_second": 206.86,
      "eval_steps_per_second": 3.319,
      "step": 4982
    },
    {
      "epoch": 48.0,
      "grad_norm": 1.2972251176834106,
      "learning_rate": 2.6000000000000002e-05,
      "loss": 0.1058,
      "step": 5088
    },
    {
      "epoch": 48.0,
      "eval_LOCATION_f1": 0.7741935483870969,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7228915662650602,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7126436781609197,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6262626262626263,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13838200271129608,
      "eval_overall_accuracy": 0.9552164123092319,
      "eval_overall_f1": 0.7951482479784365,
      "eval_overall_precision": 0.7320099255583127,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.9049,
      "eval_samples_per_second": 206.648,
      "eval_steps_per_second": 3.315,
      "step": 5088
    },
    {
      "epoch": 49.0,
      "grad_norm": 5.139357566833496,
      "learning_rate": 2.5500000000000003e-05,
      "loss": 0.1072,
      "step": 5194
    },
    {
      "epoch": 49.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7469879518072289,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6813186813186813,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8874172185430464,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8427672955974843,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6153846153846154,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5517241379310345,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1352456659078598,
      "eval_overall_accuracy": 0.9567175381536153,
      "eval_overall_f1": 0.8093278463648834,
      "eval_overall_precision": 0.7564102564102564,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.912,
      "eval_samples_per_second": 205.038,
      "eval_steps_per_second": 3.289,
      "step": 5194
    },
    {
      "epoch": 50.0,
      "grad_norm": 1.3401834964752197,
      "learning_rate": 2.5e-05,
      "loss": 0.1048,
      "step": 5300
    },
    {
      "epoch": 50.0,
      "eval_LOCATION_f1": 0.7843137254901961,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7407407407407407,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7469879518072289,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6813186813186813,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8385093167701864,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13750846683979034,
      "eval_overall_accuracy": 0.9559669752314236,
      "eval_overall_f1": 0.8076398362892225,
      "eval_overall_precision": 0.751269035532995,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9784,
      "eval_samples_per_second": 191.121,
      "eval_steps_per_second": 3.066,
      "step": 5300
    },
    {
      "epoch": 51.0,
      "grad_norm": 1.4734350442886353,
      "learning_rate": 2.45e-05,
      "loss": 0.1052,
      "step": 5406
    },
    {
      "epoch": 51.0,
      "eval_LOCATION_f1": 0.7662337662337662,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7294117647058824,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6526315789473685,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8385093167701864,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6037735849056605,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5333333333333333,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13331173360347748,
      "eval_overall_accuracy": 0.9569677257943457,
      "eval_overall_f1": 0.7989130434782608,
      "eval_overall_precision": 0.7405541561712846,
      "eval_overall_recall": 0.8672566371681416,
      "eval_runtime": 0.9147,
      "eval_samples_per_second": 204.445,
      "eval_steps_per_second": 3.28,
      "step": 5406
    },
    {
      "epoch": 52.0,
      "grad_norm": 1.5722417831420898,
      "learning_rate": 2.4e-05,
      "loss": 0.1028,
      "step": 5512
    },
    {
      "epoch": 52.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7411764705882353,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6631578947368421,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8940397350993377,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8490566037735849,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13255049288272858,
      "eval_overall_accuracy": 0.9577182887165374,
      "eval_overall_f1": 0.8043478260869565,
      "eval_overall_precision": 0.7455919395465995,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9289,
      "eval_samples_per_second": 201.318,
      "eval_steps_per_second": 3.23,
      "step": 5512
    },
    {
      "epoch": 53.0,
      "grad_norm": 2.0618319511413574,
      "learning_rate": 2.35e-05,
      "loss": 0.1022,
      "step": 5618
    },
    {
      "epoch": 53.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7294117647058824,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6526315789473685,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8844884488448846,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8375,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6037735849056605,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5333333333333333,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1327483206987381,
      "eval_overall_accuracy": 0.9564673505128847,
      "eval_overall_f1": 0.7999999999999999,
      "eval_overall_precision": 0.7424242424242424,
      "eval_overall_recall": 0.8672566371681416,
      "eval_runtime": 0.91,
      "eval_samples_per_second": 205.495,
      "eval_steps_per_second": 3.297,
      "step": 5618
    },
    {
      "epoch": 54.0,
      "grad_norm": 2.0054030418395996,
      "learning_rate": 2.3000000000000003e-05,
      "loss": 0.103,
      "step": 5724
    },
    {
      "epoch": 54.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7664670658682634,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6956521739130435,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8874172185430464,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8427672955974843,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6296296296296297,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5483870967741935,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1331961452960968,
      "eval_overall_accuracy": 0.9577182887165374,
      "eval_overall_f1": 0.8114754098360656,
      "eval_overall_precision": 0.7557251908396947,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.912,
      "eval_samples_per_second": 205.05,
      "eval_steps_per_second": 3.29,
      "step": 5724
    },
    {
      "epoch": 55.0,
      "grad_norm": 2.02776837348938,
      "learning_rate": 2.25e-05,
      "loss": 0.1018,
      "step": 5830
    },
    {
      "epoch": 55.0,
      "eval_LOCATION_f1": 0.7692307692307692,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7142857142857143,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7222222222222222,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6190476190476191,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8910891089108911,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84375,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.5964912280701754,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14076703786849976,
      "eval_overall_accuracy": 0.9542156617463098,
      "eval_overall_f1": 0.7936085219707057,
      "eval_overall_precision": 0.7233009708737864,
      "eval_overall_recall": 0.8790560471976401,
      "eval_runtime": 0.9914,
      "eval_samples_per_second": 188.627,
      "eval_steps_per_second": 3.026,
      "step": 5830
    },
    {
      "epoch": 56.0,
      "grad_norm": 1.5651822090148926,
      "learning_rate": 2.2000000000000003e-05,
      "loss": 0.1008,
      "step": 5936
    },
    {
      "epoch": 56.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7804878048780487,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7191011235955056,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8322981366459627,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13218992948532104,
      "eval_overall_accuracy": 0.9577182887165374,
      "eval_overall_f1": 0.8098495212038302,
      "eval_overall_precision": 0.7551020408163265,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9174,
      "eval_samples_per_second": 203.846,
      "eval_steps_per_second": 3.27,
      "step": 5936
    },
    {
      "epoch": 57.0,
      "grad_norm": 1.6610524654388428,
      "learning_rate": 2.15e-05,
      "loss": 0.0998,
      "step": 6042
    },
    {
      "epoch": 57.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7607361963190183,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7045454545454546,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8844884488448846,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8375,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6296296296296297,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5483870967741935,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13176319003105164,
      "eval_overall_accuracy": 0.9574681010758068,
      "eval_overall_f1": 0.8076923076923076,
      "eval_overall_precision": 0.7557840616966581,
      "eval_overall_recall": 0.8672566371681416,
      "eval_runtime": 0.9137,
      "eval_samples_per_second": 204.653,
      "eval_steps_per_second": 3.283,
      "step": 6042
    },
    {
      "epoch": 58.0,
      "grad_norm": 0.7417497038841248,
      "learning_rate": 2.1e-05,
      "loss": 0.0985,
      "step": 6148
    },
    {
      "epoch": 58.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7710843373493976,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7032967032967034,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6296296296296297,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5483870967741935,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13647660613059998,
      "eval_overall_accuracy": 0.9582186639979985,
      "eval_overall_f1": 0.8169398907103824,
      "eval_overall_precision": 0.7608142493638677,
      "eval_overall_recall": 0.8820058997050148,
      "eval_runtime": 0.9157,
      "eval_samples_per_second": 204.204,
      "eval_steps_per_second": 3.276,
      "step": 6148
    },
    {
      "epoch": 59.0,
      "grad_norm": 0.8337518572807312,
      "learning_rate": 2.05e-05,
      "loss": 0.0977,
      "step": 6254
    },
    {
      "epoch": 59.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7529411764705883,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6736842105263158,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1379348784685135,
      "eval_overall_accuracy": 0.9562171628721541,
      "eval_overall_f1": 0.8102981029810299,
      "eval_overall_precision": 0.7493734335839599,
      "eval_overall_recall": 0.8820058997050148,
      "eval_runtime": 0.9422,
      "eval_samples_per_second": 198.465,
      "eval_steps_per_second": 3.184,
      "step": 6254
    },
    {
      "epoch": 60.0,
      "grad_norm": 0.6735620498657227,
      "learning_rate": 2e-05,
      "loss": 0.0991,
      "step": 6360
    },
    {
      "epoch": 60.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7602339181286549,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6770833333333334,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13909530639648438,
      "eval_overall_accuracy": 0.9567175381536153,
      "eval_overall_f1": 0.8130081300813008,
      "eval_overall_precision": 0.7518796992481203,
      "eval_overall_recall": 0.8849557522123894,
      "eval_runtime": 0.9803,
      "eval_samples_per_second": 190.75,
      "eval_steps_per_second": 3.06,
      "step": 6360
    },
    {
      "epoch": 61.0,
      "grad_norm": 1.355182409286499,
      "learning_rate": 1.9500000000000003e-05,
      "loss": 0.0987,
      "step": 6466
    },
    {
      "epoch": 61.0,
      "eval_LOCATION_f1": 0.7741935483870969,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7228915662650602,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7428571428571429,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.65,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.14097541570663452,
      "eval_overall_accuracy": 0.9559669752314236,
      "eval_overall_f1": 0.8075370121130552,
      "eval_overall_precision": 0.7425742574257426,
      "eval_overall_recall": 0.8849557522123894,
      "eval_runtime": 0.914,
      "eval_samples_per_second": 204.604,
      "eval_steps_per_second": 3.282,
      "step": 6466
    },
    {
      "epoch": 62.0,
      "grad_norm": 0.8147560358047485,
      "learning_rate": 1.9e-05,
      "loss": 0.099,
      "step": 6572
    },
    {
      "epoch": 62.0,
      "eval_LOCATION_f1": 0.781456953642384,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7468354430379747,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7636363636363636,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8385093167701864,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13552354276180267,
      "eval_overall_accuracy": 0.9572179134350763,
      "eval_overall_f1": 0.8093278463648834,
      "eval_overall_precision": 0.7564102564102564,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.9112,
      "eval_samples_per_second": 205.226,
      "eval_steps_per_second": 3.292,
      "step": 6572
    },
    {
      "epoch": 63.0,
      "grad_norm": 0.5771262049674988,
      "learning_rate": 1.85e-05,
      "loss": 0.0967,
      "step": 6678
    },
    {
      "epoch": 63.0,
      "eval_LOCATION_f1": 0.781456953642384,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7468354430379747,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7619047619047621,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6881720430107527,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13627471029758453,
      "eval_overall_accuracy": 0.9574681010758068,
      "eval_overall_f1": 0.8125854993160055,
      "eval_overall_precision": 0.7576530612244898,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9093,
      "eval_samples_per_second": 205.662,
      "eval_steps_per_second": 3.299,
      "step": 6678
    },
    {
      "epoch": 64.0,
      "grad_norm": 1.4576925039291382,
      "learning_rate": 1.8e-05,
      "loss": 0.0976,
      "step": 6784
    },
    {
      "epoch": 64.0,
      "eval_LOCATION_f1": 0.7741935483870969,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7228915662650602,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7398843930635838,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6530612244897959,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.5714285714285715,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.48484848484848486,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13925473392009735,
      "eval_overall_accuracy": 0.9559669752314236,
      "eval_overall_f1": 0.8032345013477089,
      "eval_overall_precision": 0.739454094292804,
      "eval_overall_recall": 0.8790560471976401,
      "eval_runtime": 0.9093,
      "eval_samples_per_second": 205.657,
      "eval_steps_per_second": 3.299,
      "step": 6784
    },
    {
      "epoch": 65.0,
      "grad_norm": 12.998764991760254,
      "learning_rate": 1.75e-05,
      "loss": 0.0977,
      "step": 6890
    },
    {
      "epoch": 65.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7664670658682634,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6956521739130435,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8910891089108911,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84375,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13550983369350433,
      "eval_overall_accuracy": 0.9577182887165374,
      "eval_overall_f1": 0.8103683492496591,
      "eval_overall_precision": 0.7538071065989848,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9302,
      "eval_samples_per_second": 201.031,
      "eval_steps_per_second": 3.225,
      "step": 6890
    },
    {
      "epoch": 66.0,
      "grad_norm": 1.033444881439209,
      "learning_rate": 1.7000000000000003e-05,
      "loss": 0.096,
      "step": 6996
    },
    {
      "epoch": 66.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7619047619047621,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6881720430107527,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8910891089108911,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84375,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13633564114570618,
      "eval_overall_accuracy": 0.9574681010758068,
      "eval_overall_f1": 0.8065395095367849,
      "eval_overall_precision": 0.7493670886075949,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9065,
      "eval_samples_per_second": 206.297,
      "eval_steps_per_second": 3.31,
      "step": 6996
    },
    {
      "epoch": 67.0,
      "grad_norm": 0.6555174589157104,
      "learning_rate": 1.65e-05,
      "loss": 0.0958,
      "step": 7102
    },
    {
      "epoch": 67.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7710843373493976,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7032967032967034,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8940397350993377,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8490566037735849,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6296296296296297,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5483870967741935,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13482040166854858,
      "eval_overall_accuracy": 0.9582186639979985,
      "eval_overall_f1": 0.8136986301369864,
      "eval_overall_precision": 0.7595907928388747,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9171,
      "eval_samples_per_second": 203.894,
      "eval_steps_per_second": 3.271,
      "step": 7102
    },
    {
      "epoch": 68.0,
      "grad_norm": 1.1487456560134888,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.096,
      "step": 7208
    },
    {
      "epoch": 68.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7831325301204819,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7142857142857143,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.9006622516556291,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8553459119496856,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13745680451393127,
      "eval_overall_accuracy": 0.9582186639979985,
      "eval_overall_f1": 0.8142076502732241,
      "eval_overall_precision": 0.7582697201017812,
      "eval_overall_recall": 0.8790560471976401,
      "eval_runtime": 0.9069,
      "eval_samples_per_second": 206.198,
      "eval_steps_per_second": 3.308,
      "step": 7208
    },
    {
      "epoch": 69.0,
      "grad_norm": 1.4904353618621826,
      "learning_rate": 1.55e-05,
      "loss": 0.0956,
      "step": 7314
    },
    {
      "epoch": 69.0,
      "eval_LOCATION_f1": 0.7631578947368423,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.725,
      "eval_LOCATION_recall": 0.8055555555555556,
      "eval_ORGANIZATION_f1": 0.7590361445783131,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6923076923076923,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8940397350993377,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8490566037735849,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13514593243598938,
      "eval_overall_accuracy": 0.957968476357268,
      "eval_overall_f1": 0.8065843621399177,
      "eval_overall_precision": 0.7538461538461538,
      "eval_overall_recall": 0.8672566371681416,
      "eval_runtime": 0.9093,
      "eval_samples_per_second": 205.654,
      "eval_steps_per_second": 3.299,
      "step": 7314
    },
    {
      "epoch": 70.0,
      "grad_norm": 0.8389599919319153,
      "learning_rate": 1.5e-05,
      "loss": 0.0952,
      "step": 7420
    },
    {
      "epoch": 70.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7425149700598803,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6739130434782609,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8322981366459627,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6037735849056605,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5333333333333333,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13362941145896912,
      "eval_overall_accuracy": 0.9567175381536153,
      "eval_overall_f1": 0.8005464480874317,
      "eval_overall_precision": 0.7455470737913485,
      "eval_overall_recall": 0.8643067846607669,
      "eval_runtime": 0.9037,
      "eval_samples_per_second": 206.933,
      "eval_steps_per_second": 3.32,
      "step": 7420
    },
    {
      "epoch": 71.0,
      "grad_norm": 1.3361512422561646,
      "learning_rate": 1.45e-05,
      "loss": 0.0956,
      "step": 7526
    },
    {
      "epoch": 71.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7619047619047621,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6881720430107527,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8844884488448846,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8375,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13655199110507965,
      "eval_overall_accuracy": 0.9567175381536153,
      "eval_overall_f1": 0.8038147138964578,
      "eval_overall_precision": 0.7468354430379747,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.9117,
      "eval_samples_per_second": 205.103,
      "eval_steps_per_second": 3.29,
      "step": 7526
    },
    {
      "epoch": 72.0,
      "grad_norm": 1.1266378164291382,
      "learning_rate": 1.4000000000000001e-05,
      "loss": 0.0948,
      "step": 7632
    },
    {
      "epoch": 72.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7692307692307693,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8910891089108911,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84375,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13651823997497559,
      "eval_overall_accuracy": 0.9582186639979985,
      "eval_overall_f1": 0.8081632653061224,
      "eval_overall_precision": 0.75,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9168,
      "eval_samples_per_second": 203.972,
      "eval_steps_per_second": 3.272,
      "step": 7632
    },
    {
      "epoch": 73.0,
      "grad_norm": 1.7619799375534058,
      "learning_rate": 1.3500000000000001e-05,
      "loss": 0.096,
      "step": 7738
    },
    {
      "epoch": 73.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7619047619047621,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6881720430107527,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6037735849056605,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5333333333333333,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13442060351371765,
      "eval_overall_accuracy": 0.9574681010758068,
      "eval_overall_f1": 0.8082191780821917,
      "eval_overall_precision": 0.7544757033248082,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.9164,
      "eval_samples_per_second": 204.061,
      "eval_steps_per_second": 3.274,
      "step": 7738
    },
    {
      "epoch": 74.0,
      "grad_norm": 1.590961217880249,
      "learning_rate": 1.3000000000000001e-05,
      "loss": 0.0939,
      "step": 7844
    },
    {
      "epoch": 74.0,
      "eval_LOCATION_f1": 0.7612903225806451,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7108433734939759,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7692307692307693,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8844884488448846,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8375,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14096762239933014,
      "eval_overall_accuracy": 0.9559669752314236,
      "eval_overall_f1": 0.8005427408412482,
      "eval_overall_precision": 0.7412060301507538,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.9114,
      "eval_samples_per_second": 205.17,
      "eval_steps_per_second": 3.291,
      "step": 7844
    },
    {
      "epoch": 75.0,
      "grad_norm": 1.3473641872406006,
      "learning_rate": 1.25e-05,
      "loss": 0.0936,
      "step": 7950
    },
    {
      "epoch": 75.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7558139534883721,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6701030927835051,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8940397350993377,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8490566037735849,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13918167352676392,
      "eval_overall_accuracy": 0.9559669752314236,
      "eval_overall_f1": 0.8032564450474898,
      "eval_overall_precision": 0.7437185929648241,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9163,
      "eval_samples_per_second": 204.086,
      "eval_steps_per_second": 3.274,
      "step": 7950
    },
    {
      "epoch": 76.0,
      "grad_norm": 1.2699788808822632,
      "learning_rate": 1.2e-05,
      "loss": 0.0923,
      "step": 8056
    },
    {
      "epoch": 76.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7647058823529413,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6842105263157895,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8910891089108911,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84375,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1376781463623047,
      "eval_overall_accuracy": 0.9569677257943457,
      "eval_overall_f1": 0.8070652173913044,
      "eval_overall_precision": 0.7481108312342569,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9152,
      "eval_samples_per_second": 204.323,
      "eval_steps_per_second": 3.278,
      "step": 8056
    },
    {
      "epoch": 77.0,
      "grad_norm": 0.7005648612976074,
      "learning_rate": 1.1500000000000002e-05,
      "loss": 0.0914,
      "step": 8162
    },
    {
      "epoch": 77.0,
      "eval_LOCATION_f1": 0.7662337662337662,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7558139534883721,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6701030927835051,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13786636292934418,
      "eval_overall_accuracy": 0.9564673505128847,
      "eval_overall_f1": 0.8016304347826085,
      "eval_overall_precision": 0.743073047858942,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.9115,
      "eval_samples_per_second": 205.158,
      "eval_steps_per_second": 3.291,
      "step": 8162
    },
    {
      "epoch": 78.0,
      "grad_norm": 1.3490372896194458,
      "learning_rate": 1.1000000000000001e-05,
      "loss": 0.0921,
      "step": 8268
    },
    {
      "epoch": 78.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7515151515151515,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6888888888888889,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8940397350993377,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8490566037735849,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1341177523136139,
      "eval_overall_accuracy": 0.957968476357268,
      "eval_overall_f1": 0.8065843621399177,
      "eval_overall_precision": 0.7538461538461538,
      "eval_overall_recall": 0.8672566371681416,
      "eval_runtime": 0.9056,
      "eval_samples_per_second": 206.504,
      "eval_steps_per_second": 3.313,
      "step": 8268
    },
    {
      "epoch": 79.0,
      "grad_norm": 1.7868270874023438,
      "learning_rate": 1.05e-05,
      "loss": 0.0917,
      "step": 8374
    },
    {
      "epoch": 79.0,
      "eval_LOCATION_f1": 0.7631578947368423,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.725,
      "eval_LOCATION_recall": 0.8055555555555556,
      "eval_ORGANIZATION_f1": 0.7619047619047621,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6881720430107527,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.9,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8598726114649682,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6037735849056605,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5333333333333333,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1321086883544922,
      "eval_overall_accuracy": 0.957968476357268,
      "eval_overall_f1": 0.8104395604395604,
      "eval_overall_precision": 0.7583547557840618,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.912,
      "eval_samples_per_second": 205.039,
      "eval_steps_per_second": 3.289,
      "step": 8374
    },
    {
      "epoch": 80.0,
      "grad_norm": 0.7334626913070679,
      "learning_rate": 1e-05,
      "loss": 0.0918,
      "step": 8480
    },
    {
      "epoch": 80.0,
      "eval_LOCATION_f1": 0.7662337662337662,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7692307692307693,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8874172185430464,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8427672955974843,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1394556760787964,
      "eval_overall_accuracy": 0.9572179134350763,
      "eval_overall_f1": 0.8027210884353742,
      "eval_overall_precision": 0.7449494949494949,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.9129,
      "eval_samples_per_second": 204.847,
      "eval_steps_per_second": 3.286,
      "step": 8480
    },
    {
      "epoch": 81.0,
      "grad_norm": 1.6766005754470825,
      "learning_rate": 9.5e-06,
      "loss": 0.0925,
      "step": 8586
    },
    {
      "epoch": 81.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.757396449704142,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6808510638297872,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8874172185430464,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8427672955974843,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.137154683470726,
      "eval_overall_accuracy": 0.9572179134350763,
      "eval_overall_f1": 0.8021828103683494,
      "eval_overall_precision": 0.7461928934010152,
      "eval_overall_recall": 0.8672566371681416,
      "eval_runtime": 0.9114,
      "eval_samples_per_second": 205.177,
      "eval_steps_per_second": 3.292,
      "step": 8586
    },
    {
      "epoch": 82.0,
      "grad_norm": 1.6531654596328735,
      "learning_rate": 9e-06,
      "loss": 0.0901,
      "step": 8692
    },
    {
      "epoch": 82.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.757396449704142,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6808510638297872,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8874172185430464,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8427672955974843,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13694968819618225,
      "eval_overall_accuracy": 0.9572179134350763,
      "eval_overall_f1": 0.8021828103683494,
      "eval_overall_precision": 0.7461928934010152,
      "eval_overall_recall": 0.8672566371681416,
      "eval_runtime": 0.9646,
      "eval_samples_per_second": 193.863,
      "eval_steps_per_second": 3.11,
      "step": 8692
    },
    {
      "epoch": 83.0,
      "grad_norm": 3.4270708560943604,
      "learning_rate": 8.500000000000002e-06,
      "loss": 0.0914,
      "step": 8798
    },
    {
      "epoch": 83.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.757396449704142,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6808510638297872,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1351262480020523,
      "eval_overall_accuracy": 0.9574681010758068,
      "eval_overall_f1": 0.8032786885245901,
      "eval_overall_precision": 0.7480916030534351,
      "eval_overall_recall": 0.8672566371681416,
      "eval_runtime": 0.908,
      "eval_samples_per_second": 205.949,
      "eval_steps_per_second": 3.304,
      "step": 8798
    },
    {
      "epoch": 84.0,
      "grad_norm": 1.2858631610870361,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.0908,
      "step": 8904
    },
    {
      "epoch": 84.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7647058823529413,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6842105263157895,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13691945374011993,
      "eval_overall_accuracy": 0.9572179134350763,
      "eval_overall_f1": 0.8049113233287858,
      "eval_overall_precision": 0.748730964467005,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.9149,
      "eval_samples_per_second": 204.386,
      "eval_steps_per_second": 3.279,
      "step": 8904
    },
    {
      "epoch": 85.0,
      "grad_norm": 2.4220991134643555,
      "learning_rate": 7.5e-06,
      "loss": 0.0937,
      "step": 9010
    },
    {
      "epoch": 85.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7647058823529413,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6842105263157895,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13628537952899933,
      "eval_overall_accuracy": 0.9574681010758068,
      "eval_overall_f1": 0.8065395095367849,
      "eval_overall_precision": 0.7493670886075949,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.91,
      "eval_samples_per_second": 205.496,
      "eval_steps_per_second": 3.297,
      "step": 9010
    },
    {
      "epoch": 86.0,
      "grad_norm": 1.8708733320236206,
      "learning_rate": 7.000000000000001e-06,
      "loss": 0.0926,
      "step": 9116
    },
    {
      "epoch": 86.0,
      "eval_LOCATION_f1": 0.7662337662337662,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7380952380952381,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6666666666666666,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1368514746427536,
      "eval_overall_accuracy": 0.9569677257943457,
      "eval_overall_f1": 0.7978142076502732,
      "eval_overall_precision": 0.7430025445292621,
      "eval_overall_recall": 0.8613569321533924,
      "eval_runtime": 0.9121,
      "eval_samples_per_second": 205.021,
      "eval_steps_per_second": 3.289,
      "step": 9116
    },
    {
      "epoch": 87.0,
      "grad_norm": 1.4662730693817139,
      "learning_rate": 6.5000000000000004e-06,
      "loss": 0.09,
      "step": 9222
    },
    {
      "epoch": 87.0,
      "eval_LOCATION_f1": 0.7712418300653595,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7283950617283951,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7425149700598803,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6739130434782609,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8933333333333333,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8535031847133758,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1350514441728592,
      "eval_overall_accuracy": 0.9577182887165374,
      "eval_overall_f1": 0.803840877914952,
      "eval_overall_precision": 0.7512820512820513,
      "eval_overall_recall": 0.8643067846607669,
      "eval_runtime": 0.9064,
      "eval_samples_per_second": 206.321,
      "eval_steps_per_second": 3.31,
      "step": 9222
    },
    {
      "epoch": 88.0,
      "grad_norm": 0.9708890318870544,
      "learning_rate": 6e-06,
      "loss": 0.0895,
      "step": 9328
    },
    {
      "epoch": 88.0,
      "eval_LOCATION_f1": 0.7662337662337662,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.75,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6774193548387096,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13717986643314362,
      "eval_overall_accuracy": 0.9574681010758068,
      "eval_overall_f1": 0.8032786885245901,
      "eval_overall_precision": 0.7480916030534351,
      "eval_overall_recall": 0.8672566371681416,
      "eval_runtime": 0.9061,
      "eval_samples_per_second": 206.382,
      "eval_steps_per_second": 3.311,
      "step": 9328
    },
    {
      "epoch": 89.0,
      "grad_norm": 0.9818236231803894,
      "learning_rate": 5.500000000000001e-06,
      "loss": 0.0907,
      "step": 9434
    },
    {
      "epoch": 89.0,
      "eval_LOCATION_f1": 0.7662337662337662,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7619047619047621,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6881720430107527,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8874172185430464,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8427672955974843,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13800707459449768,
      "eval_overall_accuracy": 0.9564673505128847,
      "eval_overall_f1": 0.8010899182561309,
      "eval_overall_precision": 0.7443037974683544,
      "eval_overall_recall": 0.8672566371681416,
      "eval_runtime": 0.9188,
      "eval_samples_per_second": 203.525,
      "eval_steps_per_second": 3.265,
      "step": 9434
    },
    {
      "epoch": 90.0,
      "grad_norm": 1.852787971496582,
      "learning_rate": 5e-06,
      "loss": 0.0876,
      "step": 9540
    },
    {
      "epoch": 90.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7692307692307693,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.137681245803833,
      "eval_overall_accuracy": 0.9574681010758068,
      "eval_overall_f1": 0.8076398362892225,
      "eval_overall_precision": 0.751269035532995,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9141,
      "eval_samples_per_second": 204.564,
      "eval_steps_per_second": 3.282,
      "step": 9540
    },
    {
      "epoch": 91.0,
      "grad_norm": 1.275254487991333,
      "learning_rate": 4.5e-06,
      "loss": 0.0901,
      "step": 9646
    },
    {
      "epoch": 91.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7738095238095238,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6989247311827957,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8874172185430464,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8427672955974843,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1385199874639511,
      "eval_overall_accuracy": 0.9577182887165374,
      "eval_overall_f1": 0.8076398362892225,
      "eval_overall_precision": 0.751269035532995,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9137,
      "eval_samples_per_second": 204.659,
      "eval_steps_per_second": 3.283,
      "step": 9646
    },
    {
      "epoch": 92.0,
      "grad_norm": 6.197476387023926,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0937,
      "step": 9752
    },
    {
      "epoch": 92.0,
      "eval_LOCATION_f1": 0.7662337662337662,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.757396449704142,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6808510638297872,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1368849277496338,
      "eval_overall_accuracy": 0.9577182887165374,
      "eval_overall_f1": 0.8049113233287858,
      "eval_overall_precision": 0.748730964467005,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.9187,
      "eval_samples_per_second": 203.55,
      "eval_steps_per_second": 3.266,
      "step": 9752
    },
    {
      "epoch": 93.0,
      "grad_norm": 0.66109699010849,
      "learning_rate": 3.5000000000000004e-06,
      "loss": 0.0879,
      "step": 9858
    },
    {
      "epoch": 93.0,
      "eval_LOCATION_f1": 0.7662337662337662,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7692307692307693,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13911746442317963,
      "eval_overall_accuracy": 0.9569677257943457,
      "eval_overall_f1": 0.8038147138964578,
      "eval_overall_precision": 0.7468354430379747,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.9162,
      "eval_samples_per_second": 204.094,
      "eval_steps_per_second": 3.274,
      "step": 9858
    },
    {
      "epoch": 94.0,
      "grad_norm": 0.9965635538101196,
      "learning_rate": 3e-06,
      "loss": 0.0903,
      "step": 9964
    },
    {
      "epoch": 94.0,
      "eval_LOCATION_f1": 0.7662337662337662,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7692307692307693,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13771039247512817,
      "eval_overall_accuracy": 0.9572179134350763,
      "eval_overall_f1": 0.8065395095367849,
      "eval_overall_precision": 0.7493670886075949,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9132,
      "eval_samples_per_second": 204.777,
      "eval_steps_per_second": 3.285,
      "step": 9964
    },
    {
      "epoch": 95.0,
      "grad_norm": 0.8963160514831543,
      "learning_rate": 2.5e-06,
      "loss": 0.0895,
      "step": 10070
    },
    {
      "epoch": 95.0,
      "eval_LOCATION_f1": 0.7662337662337662,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7692307692307693,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13782991468906403,
      "eval_overall_accuracy": 0.9574681010758068,
      "eval_overall_f1": 0.8076398362892225,
      "eval_overall_precision": 0.751269035532995,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.9219,
      "eval_samples_per_second": 202.849,
      "eval_steps_per_second": 3.254,
      "step": 10070
    },
    {
      "epoch": 96.0,
      "grad_norm": 0.9854846000671387,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0923,
      "step": 10176
    },
    {
      "epoch": 96.0,
      "eval_LOCATION_f1": 0.7662337662337662,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7195121951219512,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7692307692307693,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1382056623697281,
      "eval_overall_accuracy": 0.9569677257943457,
      "eval_overall_f1": 0.8038147138964578,
      "eval_overall_precision": 0.7468354430379747,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.9084,
      "eval_samples_per_second": 205.85,
      "eval_steps_per_second": 3.302,
      "step": 10176
    },
    {
      "epoch": 97.0,
      "grad_norm": 0.8474765419960022,
      "learning_rate": 1.5e-06,
      "loss": 0.0893,
      "step": 10282
    },
    {
      "epoch": 97.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7692307692307693,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5818181818181818,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13706086575984955,
      "eval_overall_accuracy": 0.9574681010758068,
      "eval_overall_f1": 0.8092643051771118,
      "eval_overall_precision": 0.7518987341772152,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9212,
      "eval_samples_per_second": 202.999,
      "eval_steps_per_second": 3.257,
      "step": 10282
    },
    {
      "epoch": 98.0,
      "grad_norm": 2.321481704711914,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0884,
      "step": 10388
    },
    {
      "epoch": 98.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7692307692307693,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13722318410873413,
      "eval_overall_accuracy": 0.9577182887165374,
      "eval_overall_f1": 0.8103683492496591,
      "eval_overall_precision": 0.7538071065989848,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9082,
      "eval_samples_per_second": 205.909,
      "eval_steps_per_second": 3.303,
      "step": 10388
    },
    {
      "epoch": 99.0,
      "grad_norm": 0.8234521150588989,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0881,
      "step": 10494
    },
    {
      "epoch": 99.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7692307692307693,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13734257221221924,
      "eval_overall_accuracy": 0.9577182887165374,
      "eval_overall_f1": 0.8103683492496591,
      "eval_overall_precision": 0.7538071065989848,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9101,
      "eval_samples_per_second": 205.475,
      "eval_steps_per_second": 3.296,
      "step": 10494
    },
    {
      "epoch": 100.0,
      "grad_norm": 0.8098838925361633,
      "learning_rate": 0.0,
      "loss": 0.0886,
      "step": 10600
    },
    {
      "epoch": 100.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.7692307692307693,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8903654485049833,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8481012658227848,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13735896348953247,
      "eval_overall_accuracy": 0.9577182887165374,
      "eval_overall_f1": 0.8103683492496591,
      "eval_overall_precision": 0.7538071065989848,
      "eval_overall_recall": 0.8761061946902655,
      "eval_runtime": 0.9149,
      "eval_samples_per_second": 204.394,
      "eval_steps_per_second": 3.279,
      "step": 10600
    },
    {
      "epoch": 100.0,
      "step": 10600,
      "total_flos": 4517924176755654.0,
      "train_loss": 0.14866713703803297,
      "train_runtime": 1220.9124,
      "train_samples_per_second": 138.175,
      "train_steps_per_second": 8.682
    }
  ],
  "logging_steps": 500,
  "max_steps": 10600,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 100,
  "save_steps": 500,
  "total_flos": 4517924176755654.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}