trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 100.0,
  "eval_steps": 500,
  "global_step": 9600,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 1.4258437156677246,
      "learning_rate": 4.9500000000000004e-05,
      "loss": 0.8594,
      "step": 96
    },
    {
      "epoch": 1.0,
      "eval_LOCATION_f1": 0.0,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.0,
      "eval_LOCATION_recall": 0.0,
      "eval_ORGANIZATION_f1": 0.012422360248447206,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.3333333333333333,
      "eval_ORGANIZATION_recall": 0.006329113924050633,
      "eval_PERSON_f1": 0.0,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.0,
      "eval_PERSON_recall": 0.0,
      "eval_loss": 0.49973738193511963,
      "eval_overall_accuracy": 0.8396925610760362,
      "eval_overall_f1": 0.004962779156327543,
      "eval_overall_precision": 0.2,
      "eval_overall_recall": 0.002512562814070352,
      "eval_runtime": 0.3166,
      "eval_samples_per_second": 537.033,
      "eval_steps_per_second": 9.477,
      "step": 96
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.8164311051368713,
      "learning_rate": 4.9e-05,
      "loss": 0.3949,
      "step": 192
    },
    {
      "epoch": 2.0,
      "eval_LOCATION_f1": 0.4285714285714286,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.4180327868852459,
      "eval_LOCATION_recall": 0.4396551724137931,
      "eval_ORGANIZATION_f1": 0.6,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.5604395604395604,
      "eval_ORGANIZATION_recall": 0.6455696202531646,
      "eval_PERSON_f1": 0.6416382252559727,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.5562130177514792,
      "eval_PERSON_recall": 0.7580645161290323,
      "eval_loss": 0.22518467903137207,
      "eval_overall_accuracy": 0.9330222344221796,
      "eval_overall_f1": 0.5671641791044776,
      "eval_overall_precision": 0.5221987315010571,
      "eval_overall_recall": 0.6206030150753769,
      "eval_runtime": 0.3211,
      "eval_samples_per_second": 529.489,
      "eval_steps_per_second": 9.344,
      "step": 192
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.9961602091789246,
      "learning_rate": 4.85e-05,
      "loss": 0.1905,
      "step": 288
    },
    {
      "epoch": 3.0,
      "eval_LOCATION_f1": 0.7532467532467532,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.7565217391304347,
      "eval_LOCATION_recall": 0.75,
      "eval_ORGANIZATION_f1": 0.7514450867052023,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8227848101265823,
      "eval_PERSON_f1": 0.9425287356321839,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.8978102189781022,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.10961315035820007,
      "eval_overall_accuracy": 0.9654131210540763,
      "eval_overall_f1": 0.8114558472553698,
      "eval_overall_precision": 0.7727272727272727,
      "eval_overall_recall": 0.8542713567839196,
      "eval_runtime": 0.3146,
      "eval_samples_per_second": 540.424,
      "eval_steps_per_second": 9.537,
      "step": 288
    },
    {
      "epoch": 4.0,
      "grad_norm": 1.2734073400497437,
      "learning_rate": 4.8e-05,
      "loss": 0.1245,
      "step": 384
    },
    {
      "epoch": 4.0,
      "eval_LOCATION_f1": 0.804780876494024,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.7481481481481481,
      "eval_LOCATION_recall": 0.8706896551724138,
      "eval_ORGANIZATION_f1": 0.8095238095238095,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.7640449438202247,
      "eval_ORGANIZATION_recall": 0.8607594936708861,
      "eval_PERSON_f1": 0.9609375,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.9318181818181818,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.08056363463401794,
      "eval_overall_accuracy": 0.9725500960746637,
      "eval_overall_f1": 0.8540925266903914,
      "eval_overall_precision": 0.8089887640449438,
      "eval_overall_recall": 0.9045226130653267,
      "eval_runtime": 0.3159,
      "eval_samples_per_second": 538.19,
      "eval_steps_per_second": 9.497,
      "step": 384
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.53249591588974,
      "learning_rate": 4.75e-05,
      "loss": 0.1021,
      "step": 480
    },
    {
      "epoch": 5.0,
      "eval_LOCATION_f1": 0.8917748917748919,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.8956521739130435,
      "eval_LOCATION_recall": 0.8879310344827587,
      "eval_ORGANIZATION_f1": 0.8580246913580247,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.8373493975903614,
      "eval_ORGANIZATION_recall": 0.879746835443038,
      "eval_PERSON_f1": 0.9800796812749005,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.968503937007874,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05397868901491165,
      "eval_overall_accuracy": 0.9821575624485315,
      "eval_overall_f1": 0.9057071960297767,
      "eval_overall_precision": 0.8946078431372549,
      "eval_overall_recall": 0.9170854271356784,
      "eval_runtime": 0.3141,
      "eval_samples_per_second": 541.223,
      "eval_steps_per_second": 9.551,
      "step": 480
    },
    {
      "epoch": 6.0,
      "grad_norm": 1.9443389177322388,
      "learning_rate": 4.7e-05,
      "loss": 0.087,
      "step": 576
    },
    {
      "epoch": 6.0,
      "eval_LOCATION_f1": 0.8514056224899598,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.7969924812030075,
      "eval_LOCATION_recall": 0.9137931034482759,
      "eval_ORGANIZATION_f1": 0.8571428571428572,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9020979020979021,
      "eval_ORGANIZATION_recall": 0.8164556962025317,
      "eval_PERSON_f1": 0.9723320158102766,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.9534883720930233,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05299806594848633,
      "eval_overall_accuracy": 0.981059566291518,
      "eval_overall_f1": 0.8916562889165629,
      "eval_overall_precision": 0.8839506172839506,
      "eval_overall_recall": 0.8994974874371859,
      "eval_runtime": 0.3146,
      "eval_samples_per_second": 540.45,
      "eval_steps_per_second": 9.537,
      "step": 576
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.8921577334403992,
      "learning_rate": 4.6500000000000005e-05,
      "loss": 0.0748,
      "step": 672
    },
    {
      "epoch": 7.0,
      "eval_LOCATION_f1": 0.8956521739130435,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9035087719298246,
      "eval_LOCATION_recall": 0.8879310344827587,
      "eval_ORGANIZATION_f1": 0.8685015290519877,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.8402366863905325,
      "eval_ORGANIZATION_recall": 0.8987341772151899,
      "eval_PERSON_f1": 0.9840000000000001,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.9761904761904762,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.046895649284124374,
      "eval_overall_accuracy": 0.9832555586055449,
      "eval_overall_f1": 0.9120198265179678,
      "eval_overall_precision": 0.8997555012224939,
      "eval_overall_recall": 0.9246231155778895,
      "eval_runtime": 0.3181,
      "eval_samples_per_second": 534.396,
      "eval_steps_per_second": 9.431,
      "step": 672
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.35466715693473816,
      "learning_rate": 4.600000000000001e-05,
      "loss": 0.0685,
      "step": 768
    },
    {
      "epoch": 8.0,
      "eval_LOCATION_f1": 0.9191489361702128,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.907563025210084,
      "eval_LOCATION_recall": 0.9310344827586207,
      "eval_ORGANIZATION_f1": 0.9211356466876972,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9182389937106918,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9840000000000001,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.9761904761904762,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.04237267002463341,
      "eval_overall_accuracy": 0.9862750480373319,
      "eval_overall_f1": 0.9401496259351622,
      "eval_overall_precision": 0.9331683168316832,
      "eval_overall_recall": 0.9472361809045227,
      "eval_runtime": 0.3148,
      "eval_samples_per_second": 540.095,
      "eval_steps_per_second": 9.531,
      "step": 768
    },
    {
      "epoch": 9.0,
      "grad_norm": 0.43811288475990295,
      "learning_rate": 4.55e-05,
      "loss": 0.0611,
      "step": 864
    },
    {
      "epoch": 9.0,
      "eval_LOCATION_f1": 0.9159663865546218,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.8934426229508197,
      "eval_LOCATION_recall": 0.9396551724137931,
      "eval_ORGANIZATION_f1": 0.9096573208722742,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.8957055214723927,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03771281987428665,
      "eval_overall_accuracy": 0.9879220422728521,
      "eval_overall_f1": 0.9356435643564357,
      "eval_overall_precision": 0.9219512195121952,
      "eval_overall_recall": 0.949748743718593,
      "eval_runtime": 0.3156,
      "eval_samples_per_second": 538.668,
      "eval_steps_per_second": 9.506,
      "step": 864
    },
    {
      "epoch": 10.0,
      "grad_norm": 0.8697198629379272,
      "learning_rate": 4.5e-05,
      "loss": 0.0573,
      "step": 960
    },
    {
      "epoch": 10.0,
      "eval_LOCATION_f1": 0.9016393442622951,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.859375,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9102564102564102,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.922077922077922,
      "eval_ORGANIZATION_recall": 0.8987341772151899,
      "eval_PERSON_f1": 0.9840000000000001,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.9761904761904762,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.042089346796274185,
      "eval_overall_accuracy": 0.9854515509195718,
      "eval_overall_f1": 0.9305210918114144,
      "eval_overall_precision": 0.9191176470588235,
      "eval_overall_recall": 0.9422110552763819,
      "eval_runtime": 0.3158,
      "eval_samples_per_second": 538.328,
      "eval_steps_per_second": 9.5,
      "step": 960
    },
    {
      "epoch": 11.0,
      "grad_norm": 0.46367743611335754,
      "learning_rate": 4.4500000000000004e-05,
      "loss": 0.054,
      "step": 1056
    },
    {
      "epoch": 11.0,
      "eval_LOCATION_f1": 0.9198312236286921,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9008264462809917,
      "eval_LOCATION_recall": 0.9396551724137931,
      "eval_ORGANIZATION_f1": 0.932475884244373,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9477124183006536,
      "eval_ORGANIZATION_recall": 0.9177215189873418,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03589425981044769,
      "eval_overall_accuracy": 0.9884710403513588,
      "eval_overall_f1": 0.9460476787954831,
      "eval_overall_precision": 0.9448621553884712,
      "eval_overall_recall": 0.9472361809045227,
      "eval_runtime": 0.3136,
      "eval_samples_per_second": 542.106,
      "eval_steps_per_second": 9.567,
      "step": 1056
    },
    {
      "epoch": 12.0,
      "grad_norm": 1.0755535364151,
      "learning_rate": 4.4000000000000006e-05,
      "loss": 0.0454,
      "step": 1152
    },
    {
      "epoch": 12.0,
      "eval_LOCATION_f1": 0.927038626609442,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9230769230769231,
      "eval_LOCATION_recall": 0.9310344827586207,
      "eval_ORGANIZATION_f1": 0.939297124600639,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9483870967741935,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9840000000000001,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.9761904761904762,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.033515144139528275,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.949748743718593,
      "eval_overall_precision": 0.949748743718593,
      "eval_overall_recall": 0.949748743718593,
      "eval_runtime": 0.3139,
      "eval_samples_per_second": 541.553,
      "eval_steps_per_second": 9.557,
      "step": 1152
    },
    {
      "epoch": 13.0,
      "grad_norm": 0.4526771008968353,
      "learning_rate": 4.35e-05,
      "loss": 0.0426,
      "step": 1248
    },
    {
      "epoch": 13.0,
      "eval_LOCATION_f1": 0.94017094017094,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9322033898305084,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9367088607594937,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9367088607594937,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03328936547040939,
      "eval_overall_accuracy": 0.9909415317046391,
      "eval_overall_f1": 0.9536921151439299,
      "eval_overall_precision": 0.9501246882793017,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3152,
      "eval_samples_per_second": 539.282,
      "eval_steps_per_second": 9.517,
      "step": 1248
    },
    {
      "epoch": 14.0,
      "grad_norm": 0.2691395580768585,
      "learning_rate": 4.3e-05,
      "loss": 0.0392,
      "step": 1344
    },
    {
      "epoch": 14.0,
      "eval_LOCATION_f1": 0.9451476793248945,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9256198347107438,
      "eval_LOCATION_recall": 0.9655172413793104,
      "eval_ORGANIZATION_f1": 0.9299363057324842,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9358974358974359,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03476633504033089,
      "eval_overall_accuracy": 0.9895690365083722,
      "eval_overall_f1": 0.9525,
      "eval_overall_precision": 0.9477611940298507,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3182,
      "eval_samples_per_second": 534.301,
      "eval_steps_per_second": 9.429,
      "step": 1344
    },
    {
      "epoch": 15.0,
      "grad_norm": 0.7213295102119446,
      "learning_rate": 4.25e-05,
      "loss": 0.0371,
      "step": 1440
    },
    {
      "epoch": 15.0,
      "eval_LOCATION_f1": 0.9327731092436975,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9098360655737705,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9367088607594937,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9367088607594937,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.034872811287641525,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9514321295143213,
      "eval_overall_precision": 0.9432098765432099,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.3157,
      "eval_samples_per_second": 538.535,
      "eval_steps_per_second": 9.504,
      "step": 1440
    },
    {
      "epoch": 16.0,
      "grad_norm": 0.6809852123260498,
      "learning_rate": 4.2e-05,
      "loss": 0.0339,
      "step": 1536
    },
    {
      "epoch": 16.0,
      "eval_LOCATION_f1": 0.9416666666666667,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9112903225806451,
      "eval_LOCATION_recall": 0.9741379310344828,
      "eval_ORGANIZATION_f1": 0.9260450160771704,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9411764705882353,
      "eval_ORGANIZATION_recall": 0.9113924050632911,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03426952287554741,
      "eval_overall_accuracy": 0.9895690365083722,
      "eval_overall_f1": 0.95,
      "eval_overall_precision": 0.945273631840796,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3204,
      "eval_samples_per_second": 530.537,
      "eval_steps_per_second": 9.362,
      "step": 1536
    },
    {
      "epoch": 17.0,
      "grad_norm": 0.7806084156036377,
      "learning_rate": 4.15e-05,
      "loss": 0.0331,
      "step": 1632
    },
    {
      "epoch": 17.0,
      "eval_LOCATION_f1": 0.9316239316239315,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.923728813559322,
      "eval_LOCATION_recall": 0.9396551724137931,
      "eval_ORGANIZATION_f1": 0.9249999999999999,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9135802469135802,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9840000000000001,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.9761904761904762,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03590258210897446,
      "eval_overall_accuracy": 0.9895690365083722,
      "eval_overall_f1": 0.945273631840796,
      "eval_overall_precision": 0.9359605911330049,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3135,
      "eval_samples_per_second": 542.247,
      "eval_steps_per_second": 9.569,
      "step": 1632
    },
    {
      "epoch": 18.0,
      "grad_norm": 0.9416468739509583,
      "learning_rate": 4.1e-05,
      "loss": 0.031,
      "step": 1728
    },
    {
      "epoch": 18.0,
      "eval_LOCATION_f1": 0.9243697478991596,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9016393442622951,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9426751592356688,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9487179487179487,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03582395240664482,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.951310861423221,
      "eval_overall_precision": 0.9454094292803971,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3127,
      "eval_samples_per_second": 543.601,
      "eval_steps_per_second": 9.593,
      "step": 1728
    },
    {
      "epoch": 19.0,
      "grad_norm": 0.8830262422561646,
      "learning_rate": 4.05e-05,
      "loss": 0.0295,
      "step": 1824
    },
    {
      "epoch": 19.0,
      "eval_LOCATION_f1": 0.94017094017094,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9322033898305084,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9308176100628932,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.925,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03033813089132309,
      "eval_overall_accuracy": 0.9912160307438924,
      "eval_overall_f1": 0.951310861423221,
      "eval_overall_precision": 0.9454094292803971,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3211,
      "eval_samples_per_second": 529.495,
      "eval_steps_per_second": 9.344,
      "step": 1824
    },
    {
      "epoch": 20.0,
      "grad_norm": 0.6577669978141785,
      "learning_rate": 4e-05,
      "loss": 0.0273,
      "step": 1920
    },
    {
      "epoch": 20.0,
      "eval_LOCATION_f1": 0.9411764705882353,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9180327868852459,
      "eval_LOCATION_recall": 0.9655172413793104,
      "eval_ORGANIZATION_f1": 0.926984126984127,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9299363057324841,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.034174975007772446,
      "eval_overall_accuracy": 0.9906670326653857,
      "eval_overall_f1": 0.9501246882793017,
      "eval_overall_precision": 0.943069306930693,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3146,
      "eval_samples_per_second": 540.438,
      "eval_steps_per_second": 9.537,
      "step": 1920
    },
    {
      "epoch": 21.0,
      "grad_norm": 0.6675641536712646,
      "learning_rate": 3.9500000000000005e-05,
      "loss": 0.0258,
      "step": 2016
    },
    {
      "epoch": 21.0,
      "eval_LOCATION_f1": 0.9256198347107438,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.8888888888888888,
      "eval_LOCATION_recall": 0.9655172413793104,
      "eval_ORGANIZATION_f1": 0.9333333333333332,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9363057324840764,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03838442638516426,
      "eval_overall_accuracy": 0.9890200384298655,
      "eval_overall_f1": 0.9478908188585609,
      "eval_overall_precision": 0.9362745098039216,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.3158,
      "eval_samples_per_second": 538.345,
      "eval_steps_per_second": 9.5,
      "step": 2016
    },
    {
      "epoch": 22.0,
      "grad_norm": 0.4553017020225525,
      "learning_rate": 3.9000000000000006e-05,
      "loss": 0.0225,
      "step": 2112
    },
    {
      "epoch": 22.0,
      "eval_LOCATION_f1": 0.9316239316239315,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.923728813559322,
      "eval_LOCATION_recall": 0.9396551724137931,
      "eval_ORGANIZATION_f1": 0.9341692789968652,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9254658385093167,
      "eval_ORGANIZATION_recall": 0.9430379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.032226499170064926,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9501246882793017,
      "eval_overall_precision": 0.943069306930693,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3168,
      "eval_samples_per_second": 536.612,
      "eval_steps_per_second": 9.47,
      "step": 2112
    },
    {
      "epoch": 23.0,
      "grad_norm": 0.5542384386062622,
      "learning_rate": 3.85e-05,
      "loss": 0.0211,
      "step": 2208
    },
    {
      "epoch": 23.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.939297124600639,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9483870967741935,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9840000000000001,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.9761904761904762,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03527655825018883,
      "eval_overall_accuracy": 0.9903925336261323,
      "eval_overall_f1": 0.9523809523809523,
      "eval_overall_precision": 0.95,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3151,
      "eval_samples_per_second": 539.433,
      "eval_steps_per_second": 9.519,
      "step": 2208
    },
    {
      "epoch": 24.0,
      "grad_norm": 0.1748531609773636,
      "learning_rate": 3.8e-05,
      "loss": 0.0213,
      "step": 2304
    },
    {
      "epoch": 24.0,
      "eval_LOCATION_f1": 0.944206008583691,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9401709401709402,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9294871794871796,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9415584415584416,
      "eval_ORGANIZATION_recall": 0.9177215189873418,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03550584986805916,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9521410579345088,
      "eval_overall_precision": 0.9545454545454546,
      "eval_overall_recall": 0.949748743718593,
      "eval_runtime": 0.315,
      "eval_samples_per_second": 539.732,
      "eval_steps_per_second": 9.525,
      "step": 2304
    },
    {
      "epoch": 25.0,
      "grad_norm": 1.2292771339416504,
      "learning_rate": 3.7500000000000003e-05,
      "loss": 0.0197,
      "step": 2400
    },
    {
      "epoch": 25.0,
      "eval_LOCATION_f1": 0.9367088607594937,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9173553719008265,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9423076923076923,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9545454545454546,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03419380262494087,
      "eval_overall_accuracy": 0.9909415317046391,
      "eval_overall_f1": 0.9548872180451128,
      "eval_overall_precision": 0.9525,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3151,
      "eval_samples_per_second": 539.494,
      "eval_steps_per_second": 9.52,
      "step": 2400
    },
    {
      "epoch": 26.0,
      "grad_norm": 0.6817178726196289,
      "learning_rate": 3.7e-05,
      "loss": 0.022,
      "step": 2496
    },
    {
      "epoch": 26.0,
      "eval_LOCATION_f1": 0.9356223175965666,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9316239316239316,
      "eval_LOCATION_recall": 0.9396551724137931,
      "eval_ORGANIZATION_f1": 0.9358974358974359,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.948051948051948,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03494803607463837,
      "eval_overall_accuracy": 0.9903925336261323,
      "eval_overall_f1": 0.9521410579345088,
      "eval_overall_precision": 0.9545454545454546,
      "eval_overall_recall": 0.949748743718593,
      "eval_runtime": 0.3145,
      "eval_samples_per_second": 540.621,
      "eval_steps_per_second": 9.54,
      "step": 2496
    },
    {
      "epoch": 27.0,
      "grad_norm": 0.8213551044464111,
      "learning_rate": 3.65e-05,
      "loss": 0.019,
      "step": 2592
    },
    {
      "epoch": 27.0,
      "eval_LOCATION_f1": 0.944206008583691,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9401709401709402,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9430379746835443,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9430379746835443,
      "eval_ORGANIZATION_recall": 0.9430379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.03204932063817978,
      "eval_overall_accuracy": 0.9912160307438924,
      "eval_overall_f1": 0.9573934837092732,
      "eval_overall_precision": 0.955,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.3151,
      "eval_samples_per_second": 539.515,
      "eval_steps_per_second": 9.521,
      "step": 2592
    },
    {
      "epoch": 28.0,
      "grad_norm": 0.803354799747467,
      "learning_rate": 3.6e-05,
      "loss": 0.0174,
      "step": 2688
    },
    {
      "epoch": 28.0,
      "eval_LOCATION_f1": 0.9572649572649573,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9491525423728814,
      "eval_LOCATION_recall": 0.9655172413793104,
      "eval_ORGANIZATION_f1": 0.939297124600639,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9483870967741935,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.041541844606399536,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9597989949748744,
      "eval_overall_precision": 0.9597989949748744,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.3143,
      "eval_samples_per_second": 540.856,
      "eval_steps_per_second": 9.545,
      "step": 2688
    },
    {
      "epoch": 29.0,
      "grad_norm": 0.15000270307064056,
      "learning_rate": 3.55e-05,
      "loss": 0.0164,
      "step": 2784
    },
    {
      "epoch": 29.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9337539432176657,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9308176100628931,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.038267266005277634,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9526184538653367,
      "eval_overall_precision": 0.9455445544554455,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.3143,
      "eval_samples_per_second": 540.813,
      "eval_steps_per_second": 9.544,
      "step": 2784
    },
    {
      "epoch": 30.0,
      "grad_norm": 0.7412194609642029,
      "learning_rate": 3.5e-05,
      "loss": 0.0166,
      "step": 2880
    },
    {
      "epoch": 30.0,
      "eval_LOCATION_f1": 0.94017094017094,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9322033898305084,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9141104294478527,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.8869047619047619,
      "eval_ORGANIZATION_recall": 0.9430379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.041352592408657074,
      "eval_overall_accuracy": 0.9890200384298655,
      "eval_overall_f1": 0.9443757725587145,
      "eval_overall_precision": 0.9294403892944039,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.3163,
      "eval_samples_per_second": 537.459,
      "eval_steps_per_second": 9.485,
      "step": 2880
    },
    {
      "epoch": 31.0,
      "grad_norm": 0.09500475227832794,
      "learning_rate": 3.45e-05,
      "loss": 0.0162,
      "step": 2976
    },
    {
      "epoch": 31.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9333333333333332,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9363057324840764,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.037069208920001984,
      "eval_overall_accuracy": 0.9906670326653857,
      "eval_overall_f1": 0.9525,
      "eval_overall_precision": 0.9477611940298507,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3185,
      "eval_samples_per_second": 533.732,
      "eval_steps_per_second": 9.419,
      "step": 2976
    },
    {
      "epoch": 32.0,
      "grad_norm": 0.6351097822189331,
      "learning_rate": 3.4000000000000007e-05,
      "loss": 0.0157,
      "step": 3072
    },
    {
      "epoch": 32.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9329073482428115,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9419354838709677,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9840000000000001,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.9761904761904762,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.04019623249769211,
      "eval_overall_accuracy": 0.9892945374691189,
      "eval_overall_f1": 0.951188986232791,
      "eval_overall_precision": 0.9476309226932669,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3128,
      "eval_samples_per_second": 543.413,
      "eval_steps_per_second": 9.59,
      "step": 3072
    },
    {
      "epoch": 33.0,
      "grad_norm": 0.40506356954574585,
      "learning_rate": 3.35e-05,
      "loss": 0.0137,
      "step": 3168
    },
    {
      "epoch": 33.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9354838709677419,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9539473684210527,
      "eval_ORGANIZATION_recall": 0.9177215189873418,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.04279535636305809,
      "eval_overall_accuracy": 0.9892945374691189,
      "eval_overall_f1": 0.9534591194968555,
      "eval_overall_precision": 0.9546599496221663,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3134,
      "eval_samples_per_second": 542.41,
      "eval_steps_per_second": 9.572,
      "step": 3168
    },
    {
      "epoch": 34.0,
      "grad_norm": 0.6109006404876709,
      "learning_rate": 3.3e-05,
      "loss": 0.0146,
      "step": 3264
    },
    {
      "epoch": 34.0,
      "eval_LOCATION_f1": 0.9491525423728815,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9333333333333333,
      "eval_LOCATION_recall": 0.9655172413793104,
      "eval_ORGANIZATION_f1": 0.9333333333333332,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9363057324840764,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.04216236621141434,
      "eval_overall_accuracy": 0.9895690365083722,
      "eval_overall_f1": 0.9550000000000001,
      "eval_overall_precision": 0.9502487562189055,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.3164,
      "eval_samples_per_second": 537.331,
      "eval_steps_per_second": 9.482,
      "step": 3264
    },
    {
      "epoch": 35.0,
      "grad_norm": 1.0238646268844604,
      "learning_rate": 3.2500000000000004e-05,
      "loss": 0.0139,
      "step": 3360
    },
    {
      "epoch": 35.0,
      "eval_LOCATION_f1": 0.9491525423728815,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9333333333333333,
      "eval_LOCATION_recall": 0.9655172413793104,
      "eval_ORGANIZATION_f1": 0.926984126984127,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9299363057324841,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9799196787148594,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.976,
      "eval_PERSON_recall": 0.9838709677419355,
      "eval_loss": 0.043526869267225266,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.95,
      "eval_overall_precision": 0.945273631840796,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.314,
      "eval_samples_per_second": 541.364,
      "eval_steps_per_second": 9.553,
      "step": 3360
    },
    {
      "epoch": 36.0,
      "grad_norm": 0.3887651860713959,
      "learning_rate": 3.2000000000000005e-05,
      "loss": 0.0126,
      "step": 3456
    },
    {
      "epoch": 36.0,
      "eval_LOCATION_f1": 0.9446808510638298,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9327731092436975,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9449838187702266,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9668874172185431,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.04276629164814949,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9583858764186632,
      "eval_overall_precision": 0.9620253164556962,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.313,
      "eval_samples_per_second": 543.059,
      "eval_steps_per_second": 9.583,
      "step": 3456
    },
    {
      "epoch": 37.0,
      "grad_norm": 1.1276990175247192,
      "learning_rate": 3.15e-05,
      "loss": 0.0112,
      "step": 3552
    },
    {
      "epoch": 37.0,
      "eval_LOCATION_f1": 0.9535864978902953,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9338842975206612,
      "eval_LOCATION_recall": 0.9741379310344828,
      "eval_ORGANIZATION_f1": 0.9423076923076923,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9545454545454546,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.04038548842072487,
      "eval_overall_accuracy": 0.9912160307438924,
      "eval_overall_f1": 0.9598997493734336,
      "eval_overall_precision": 0.9575,
      "eval_overall_recall": 0.9623115577889447,
      "eval_runtime": 0.3137,
      "eval_samples_per_second": 541.927,
      "eval_steps_per_second": 9.563,
      "step": 3552
    },
    {
      "epoch": 38.0,
      "grad_norm": 0.35374054312705994,
      "learning_rate": 3.1e-05,
      "loss": 0.0121,
      "step": 3648
    },
    {
      "epoch": 38.0,
      "eval_LOCATION_f1": 0.9446808510638298,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9327731092436975,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9446254071661238,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9731543624161074,
      "eval_ORGANIZATION_recall": 0.9177215189873418,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.04826612398028374,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9582806573957017,
      "eval_overall_precision": 0.9643765903307888,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3119,
      "eval_samples_per_second": 545.086,
      "eval_steps_per_second": 9.619,
      "step": 3648
    },
    {
      "epoch": 39.0,
      "grad_norm": 0.3826542794704437,
      "learning_rate": 3.05e-05,
      "loss": 0.0101,
      "step": 3744
    },
    {
      "epoch": 39.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9415584415584415,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9666666666666667,
      "eval_ORGANIZATION_recall": 0.9177215189873418,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.046513769775629044,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9558638083228247,
      "eval_overall_precision": 0.959493670886076,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3177,
      "eval_samples_per_second": 535.158,
      "eval_steps_per_second": 9.444,
      "step": 3744
    },
    {
      "epoch": 40.0,
      "grad_norm": 0.4108656346797943,
      "learning_rate": 3e-05,
      "loss": 0.0106,
      "step": 3840
    },
    {
      "epoch": 40.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9415584415584415,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9666666666666667,
      "eval_ORGANIZATION_recall": 0.9177215189873418,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.043700963258743286,
      "eval_overall_accuracy": 0.9895690365083722,
      "eval_overall_f1": 0.9545454545454546,
      "eval_overall_precision": 0.9593908629441624,
      "eval_overall_recall": 0.949748743718593,
      "eval_runtime": 0.3162,
      "eval_samples_per_second": 537.651,
      "eval_steps_per_second": 9.488,
      "step": 3840
    },
    {
      "epoch": 41.0,
      "grad_norm": 0.023255372419953346,
      "learning_rate": 2.95e-05,
      "loss": 0.0113,
      "step": 3936
    },
    {
      "epoch": 41.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.926984126984127,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9299363057324841,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.047751251608133316,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.95,
      "eval_overall_precision": 0.945273631840796,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3144,
      "eval_samples_per_second": 540.695,
      "eval_steps_per_second": 9.542,
      "step": 3936
    },
    {
      "epoch": 42.0,
      "grad_norm": 0.18263016641139984,
      "learning_rate": 2.9e-05,
      "loss": 0.0088,
      "step": 4032
    },
    {
      "epoch": 42.0,
      "eval_LOCATION_f1": 0.9327731092436975,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9098360655737705,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9389067524115756,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.954248366013072,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05291637405753136,
      "eval_overall_accuracy": 0.9887455393906122,
      "eval_overall_f1": 0.9523809523809523,
      "eval_overall_precision": 0.95,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3137,
      "eval_samples_per_second": 541.863,
      "eval_steps_per_second": 9.562,
      "step": 4032
    },
    {
      "epoch": 43.0,
      "grad_norm": 0.17454615235328674,
      "learning_rate": 2.8499999999999998e-05,
      "loss": 0.0092,
      "step": 4128
    },
    {
      "epoch": 43.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9381107491856678,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9664429530201343,
      "eval_ORGANIZATION_recall": 0.9113924050632911,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.049351226538419724,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9545454545454546,
      "eval_overall_precision": 0.9593908629441624,
      "eval_overall_recall": 0.949748743718593,
      "eval_runtime": 0.312,
      "eval_samples_per_second": 544.95,
      "eval_steps_per_second": 9.617,
      "step": 4128
    },
    {
      "epoch": 44.0,
      "grad_norm": 0.06936214119195938,
      "learning_rate": 2.8000000000000003e-05,
      "loss": 0.0083,
      "step": 4224
    },
    {
      "epoch": 44.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9514563106796117,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9735099337748344,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.04810861870646477,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9596977329974811,
      "eval_overall_precision": 0.9621212121212122,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3154,
      "eval_samples_per_second": 539.079,
      "eval_steps_per_second": 9.513,
      "step": 4224
    },
    {
      "epoch": 45.0,
      "grad_norm": 0.9196687340736389,
      "learning_rate": 2.7500000000000004e-05,
      "loss": 0.0104,
      "step": 4320
    },
    {
      "epoch": 45.0,
      "eval_LOCATION_f1": 0.9367088607594937,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9173553719008265,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9381107491856678,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9664429530201343,
      "eval_ORGANIZATION_recall": 0.9113924050632911,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05476151779294014,
      "eval_overall_accuracy": 0.9890200384298655,
      "eval_overall_f1": 0.953341740226986,
      "eval_overall_precision": 0.9569620253164557,
      "eval_overall_recall": 0.949748743718593,
      "eval_runtime": 0.3151,
      "eval_samples_per_second": 539.575,
      "eval_steps_per_second": 9.522,
      "step": 4320
    },
    {
      "epoch": 46.0,
      "grad_norm": 0.12706241011619568,
      "learning_rate": 2.7000000000000002e-05,
      "loss": 0.0085,
      "step": 4416
    },
    {
      "epoch": 46.0,
      "eval_LOCATION_f1": 0.9327731092436975,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9098360655737705,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9487179487179488,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.961038961038961,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05352696776390076,
      "eval_overall_accuracy": 0.9895690365083722,
      "eval_overall_f1": 0.9561952440550688,
      "eval_overall_precision": 0.9526184538653366,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.3133,
      "eval_samples_per_second": 542.597,
      "eval_steps_per_second": 9.575,
      "step": 4416
    },
    {
      "epoch": 47.0,
      "grad_norm": 0.06963124871253967,
      "learning_rate": 2.6500000000000004e-05,
      "loss": 0.0075,
      "step": 4512
    },
    {
      "epoch": 47.0,
      "eval_LOCATION_f1": 0.9282700421940928,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9090909090909091,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9419354838709677,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9605263157894737,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9799196787148594,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.976,
      "eval_PERSON_recall": 0.9838709677419355,
      "eval_loss": 0.05333380028605461,
      "eval_overall_accuracy": 0.9887455393906122,
      "eval_overall_f1": 0.949748743718593,
      "eval_overall_precision": 0.949748743718593,
      "eval_overall_recall": 0.949748743718593,
      "eval_runtime": 0.3128,
      "eval_samples_per_second": 543.415,
      "eval_steps_per_second": 9.59,
      "step": 4512
    },
    {
      "epoch": 48.0,
      "grad_norm": 0.5458090901374817,
      "learning_rate": 2.6000000000000002e-05,
      "loss": 0.0084,
      "step": 4608
    },
    {
      "epoch": 48.0,
      "eval_LOCATION_f1": 0.9491525423728815,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9333333333333333,
      "eval_LOCATION_recall": 0.9655172413793104,
      "eval_ORGANIZATION_f1": 0.9456869009584665,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9548387096774194,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.04624834656715393,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9598997493734336,
      "eval_overall_precision": 0.9575,
      "eval_overall_recall": 0.9623115577889447,
      "eval_runtime": 0.3113,
      "eval_samples_per_second": 546.028,
      "eval_steps_per_second": 9.636,
      "step": 4608
    },
    {
      "epoch": 49.0,
      "grad_norm": 0.19796976447105408,
      "learning_rate": 2.5500000000000003e-05,
      "loss": 0.007,
      "step": 4704
    },
    {
      "epoch": 49.0,
      "eval_LOCATION_f1": 0.9446808510638298,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9327731092436975,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9514563106796117,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9735099337748344,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05179499462246895,
      "eval_overall_accuracy": 0.9906670326653857,
      "eval_overall_f1": 0.9609079445145018,
      "eval_overall_precision": 0.9645569620253165,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3122,
      "eval_samples_per_second": 544.472,
      "eval_steps_per_second": 9.608,
      "step": 4704
    },
    {
      "epoch": 50.0,
      "grad_norm": 0.05334031581878662,
      "learning_rate": 2.5e-05,
      "loss": 0.0073,
      "step": 4800
    },
    {
      "epoch": 50.0,
      "eval_LOCATION_f1": 0.9491525423728815,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9333333333333333,
      "eval_LOCATION_recall": 0.9655172413793104,
      "eval_ORGANIZATION_f1": 0.9487179487179488,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.961038961038961,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.046464741230010986,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9611041405269761,
      "eval_overall_precision": 0.9598997493734336,
      "eval_overall_recall": 0.9623115577889447,
      "eval_runtime": 0.3089,
      "eval_samples_per_second": 550.262,
      "eval_steps_per_second": 9.711,
      "step": 4800
    },
    {
      "epoch": 51.0,
      "grad_norm": 0.6792040467262268,
      "learning_rate": 2.45e-05,
      "loss": 0.0072,
      "step": 4896
    },
    {
      "epoch": 51.0,
      "eval_LOCATION_f1": 0.9451476793248945,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9256198347107438,
      "eval_LOCATION_recall": 0.9655172413793104,
      "eval_ORGANIZATION_f1": 0.9517684887459807,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9673202614379085,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05347295477986336,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9611041405269761,
      "eval_overall_precision": 0.9598997493734336,
      "eval_overall_recall": 0.9623115577889447,
      "eval_runtime": 0.3094,
      "eval_samples_per_second": 549.442,
      "eval_steps_per_second": 9.696,
      "step": 4896
    },
    {
      "epoch": 52.0,
      "grad_norm": 1.0405575037002563,
      "learning_rate": 2.4e-05,
      "loss": 0.0066,
      "step": 4992
    },
    {
      "epoch": 52.0,
      "eval_LOCATION_f1": 0.9327731092436975,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9098360655737705,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9367088607594937,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9367088607594937,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05243299528956413,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9514321295143213,
      "eval_overall_precision": 0.9432098765432099,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.3116,
      "eval_samples_per_second": 545.634,
      "eval_steps_per_second": 9.629,
      "step": 4992
    },
    {
      "epoch": 53.0,
      "grad_norm": 0.7544435858726501,
      "learning_rate": 2.35e-05,
      "loss": 0.0075,
      "step": 5088
    },
    {
      "epoch": 53.0,
      "eval_LOCATION_f1": 0.9451476793248945,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9256198347107438,
      "eval_LOCATION_recall": 0.9655172413793104,
      "eval_ORGANIZATION_f1": 0.9483870967741935,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9671052631578947,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.053470365703105927,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9597989949748744,
      "eval_overall_precision": 0.9597989949748744,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.31,
      "eval_samples_per_second": 548.349,
      "eval_steps_per_second": 9.677,
      "step": 5088
    },
    {
      "epoch": 54.0,
      "grad_norm": 0.3421597182750702,
      "learning_rate": 2.3000000000000003e-05,
      "loss": 0.0068,
      "step": 5184
    },
    {
      "epoch": 54.0,
      "eval_LOCATION_f1": 0.9367088607594937,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9173553719008265,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9456869009584665,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9548387096774194,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05892359837889671,
      "eval_overall_accuracy": 0.9895690365083722,
      "eval_overall_f1": 0.9561952440550688,
      "eval_overall_precision": 0.9526184538653366,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.313,
      "eval_samples_per_second": 543.172,
      "eval_steps_per_second": 9.585,
      "step": 5184
    },
    {
      "epoch": 55.0,
      "grad_norm": 0.4127174913883209,
      "learning_rate": 2.25e-05,
      "loss": 0.007,
      "step": 5280
    },
    {
      "epoch": 55.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9456869009584665,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9548387096774194,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.04972882568836212,
      "eval_overall_accuracy": 0.9909415317046391,
      "eval_overall_f1": 0.9573934837092732,
      "eval_overall_precision": 0.955,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.31,
      "eval_samples_per_second": 548.397,
      "eval_steps_per_second": 9.678,
      "step": 5280
    },
    {
      "epoch": 56.0,
      "grad_norm": 0.4509964883327484,
      "learning_rate": 2.2000000000000003e-05,
      "loss": 0.0063,
      "step": 5376
    },
    {
      "epoch": 56.0,
      "eval_LOCATION_f1": 0.9451476793248945,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9256198347107438,
      "eval_LOCATION_recall": 0.9655172413793104,
      "eval_ORGANIZATION_f1": 0.9371069182389938,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.93125,
      "eval_ORGANIZATION_recall": 0.9430379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05140337720513344,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9552238805970149,
      "eval_overall_precision": 0.9458128078817734,
      "eval_overall_recall": 0.964824120603015,
      "eval_runtime": 0.3108,
      "eval_samples_per_second": 547.022,
      "eval_steps_per_second": 9.653,
      "step": 5376
    },
    {
      "epoch": 57.0,
      "grad_norm": 0.7318273782730103,
      "learning_rate": 2.15e-05,
      "loss": 0.0051,
      "step": 5472
    },
    {
      "epoch": 57.0,
      "eval_LOCATION_f1": 0.9446808510638298,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9327731092436975,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9456869009584665,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9548387096774194,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.052676744759082794,
      "eval_overall_accuracy": 0.9909415317046391,
      "eval_overall_f1": 0.958594730238394,
      "eval_overall_precision": 0.9573934837092731,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.3106,
      "eval_samples_per_second": 547.313,
      "eval_steps_per_second": 9.658,
      "step": 5472
    },
    {
      "epoch": 58.0,
      "grad_norm": 0.26017698645591736,
      "learning_rate": 2.1e-05,
      "loss": 0.0064,
      "step": 5568
    },
    {
      "epoch": 58.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9456869009584665,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9548387096774194,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.056647952646017075,
      "eval_overall_accuracy": 0.9906670326653857,
      "eval_overall_f1": 0.9573934837092732,
      "eval_overall_precision": 0.955,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.3115,
      "eval_samples_per_second": 545.724,
      "eval_steps_per_second": 9.63,
      "step": 5568
    },
    {
      "epoch": 59.0,
      "grad_norm": 0.08687864243984222,
      "learning_rate": 2.05e-05,
      "loss": 0.0049,
      "step": 5664
    },
    {
      "epoch": 59.0,
      "eval_LOCATION_f1": 0.9367088607594937,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9173553719008265,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.930379746835443,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.930379746835443,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05729100853204727,
      "eval_overall_accuracy": 0.9890200384298655,
      "eval_overall_f1": 0.9501246882793017,
      "eval_overall_precision": 0.943069306930693,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.31,
      "eval_samples_per_second": 548.425,
      "eval_steps_per_second": 9.678,
      "step": 5664
    },
    {
      "epoch": 60.0,
      "grad_norm": 0.07127294689416885,
      "learning_rate": 2e-05,
      "loss": 0.0046,
      "step": 5760
    },
    {
      "epoch": 60.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9426751592356688,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9487179487179487,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05765029788017273,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9561952440550688,
      "eval_overall_precision": 0.9526184538653366,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.31,
      "eval_samples_per_second": 548.423,
      "eval_steps_per_second": 9.678,
      "step": 5760
    },
    {
      "epoch": 61.0,
      "grad_norm": 1.0418661832809448,
      "learning_rate": 1.9500000000000003e-05,
      "loss": 0.0049,
      "step": 5856
    },
    {
      "epoch": 61.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9389067524115756,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.954248366013072,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9799196787148594,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.976,
      "eval_PERSON_recall": 0.9838709677419355,
      "eval_loss": 0.05813300237059593,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9522613065326633,
      "eval_overall_precision": 0.9522613065326633,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3111,
      "eval_samples_per_second": 546.389,
      "eval_steps_per_second": 9.642,
      "step": 5856
    },
    {
      "epoch": 62.0,
      "grad_norm": 0.0857049897313118,
      "learning_rate": 1.9e-05,
      "loss": 0.0043,
      "step": 5952
    },
    {
      "epoch": 62.0,
      "eval_LOCATION_f1": 0.9446808510638298,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9327731092436975,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9456869009584665,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9548387096774194,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9799196787148594,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.976,
      "eval_PERSON_recall": 0.9838709677419355,
      "eval_loss": 0.05701900273561478,
      "eval_overall_accuracy": 0.9903925336261323,
      "eval_overall_f1": 0.9560853199498118,
      "eval_overall_precision": 0.9548872180451128,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3112,
      "eval_samples_per_second": 546.27,
      "eval_steps_per_second": 9.64,
      "step": 5952
    },
    {
      "epoch": 63.0,
      "grad_norm": 0.03378366678953171,
      "learning_rate": 1.85e-05,
      "loss": 0.0048,
      "step": 6048
    },
    {
      "epoch": 63.0,
      "eval_LOCATION_f1": 0.9322033898305084,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9166666666666666,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9456869009584665,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9548387096774194,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05717466026544571,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9548872180451128,
      "eval_overall_precision": 0.9525,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3124,
      "eval_samples_per_second": 544.191,
      "eval_steps_per_second": 9.603,
      "step": 6048
    },
    {
      "epoch": 64.0,
      "grad_norm": 0.06232772022485733,
      "learning_rate": 1.8e-05,
      "loss": 0.0055,
      "step": 6144
    },
    {
      "epoch": 64.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.939297124600639,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9483870967741935,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05859093368053436,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9548872180451128,
      "eval_overall_precision": 0.9525,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3094,
      "eval_samples_per_second": 549.445,
      "eval_steps_per_second": 9.696,
      "step": 6144
    },
    {
      "epoch": 65.0,
      "grad_norm": 0.9216206669807434,
      "learning_rate": 1.75e-05,
      "loss": 0.0052,
      "step": 6240
    },
    {
      "epoch": 65.0,
      "eval_LOCATION_f1": 0.9322033898305084,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9166666666666666,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9426751592356688,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9487179487179487,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05924278870224953,
      "eval_overall_accuracy": 0.9903925336261323,
      "eval_overall_f1": 0.9536921151439299,
      "eval_overall_precision": 0.9501246882793017,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3148,
      "eval_samples_per_second": 539.974,
      "eval_steps_per_second": 9.529,
      "step": 6240
    },
    {
      "epoch": 66.0,
      "grad_norm": 0.8325523138046265,
      "learning_rate": 1.7000000000000003e-05,
      "loss": 0.0041,
      "step": 6336
    },
    {
      "epoch": 66.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9363057324840764,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9423076923076923,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9799196787148594,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.976,
      "eval_PERSON_recall": 0.9838709677419355,
      "eval_loss": 0.05509481951594353,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.951188986232791,
      "eval_overall_precision": 0.9476309226932669,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3099,
      "eval_samples_per_second": 548.64,
      "eval_steps_per_second": 9.682,
      "step": 6336
    },
    {
      "epoch": 67.0,
      "grad_norm": 0.14074504375457764,
      "learning_rate": 1.65e-05,
      "loss": 0.0048,
      "step": 6432
    },
    {
      "epoch": 67.0,
      "eval_LOCATION_f1": 0.9367088607594937,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9173553719008265,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9385113268608414,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9602649006622517,
      "eval_ORGANIZATION_recall": 0.9177215189873418,
      "eval_PERSON_f1": 0.9799196787148594,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.976,
      "eval_PERSON_recall": 0.9838709677419355,
      "eval_loss": 0.0619182251393795,
      "eval_overall_accuracy": 0.9890200384298655,
      "eval_overall_f1": 0.950943396226415,
      "eval_overall_precision": 0.9521410579345088,
      "eval_overall_recall": 0.949748743718593,
      "eval_runtime": 0.3112,
      "eval_samples_per_second": 546.197,
      "eval_steps_per_second": 9.639,
      "step": 6432
    },
    {
      "epoch": 68.0,
      "grad_norm": 0.7965418696403503,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.0033,
      "step": 6528
    },
    {
      "epoch": 68.0,
      "eval_LOCATION_f1": 0.9322033898305084,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9166666666666666,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9396825396825397,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9426751592356688,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05904194340109825,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9525,
      "eval_overall_precision": 0.9477611940298507,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.312,
      "eval_samples_per_second": 544.865,
      "eval_steps_per_second": 9.615,
      "step": 6528
    },
    {
      "epoch": 69.0,
      "grad_norm": 0.01660473830997944,
      "learning_rate": 1.55e-05,
      "loss": 0.0038,
      "step": 6624
    },
    {
      "epoch": 69.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9453376205787781,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9607843137254902,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.058813340961933136,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.957286432160804,
      "eval_overall_precision": 0.957286432160804,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3096,
      "eval_samples_per_second": 549.058,
      "eval_steps_per_second": 9.689,
      "step": 6624
    },
    {
      "epoch": 70.0,
      "grad_norm": 0.10545208305120468,
      "learning_rate": 1.5e-05,
      "loss": 0.0043,
      "step": 6720
    },
    {
      "epoch": 70.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9308176100628932,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.925,
      "eval_ORGANIZATION_recall": 0.9367088607594937,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.0531315915286541,
      "eval_overall_accuracy": 0.9895690365083722,
      "eval_overall_f1": 0.9514321295143213,
      "eval_overall_precision": 0.9432098765432099,
      "eval_overall_recall": 0.9597989949748744,
      "eval_runtime": 0.3113,
      "eval_samples_per_second": 546.059,
      "eval_steps_per_second": 9.636,
      "step": 6720
    },
    {
      "epoch": 71.0,
      "grad_norm": 0.045872997492551804,
      "learning_rate": 1.45e-05,
      "loss": 0.0051,
      "step": 6816
    },
    {
      "epoch": 71.0,
      "eval_LOCATION_f1": 0.9237288135593221,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9083333333333333,
      "eval_LOCATION_recall": 0.9396551724137931,
      "eval_ORGANIZATION_f1": 0.930379746835443,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.930379746835443,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.057702187448740005,
      "eval_overall_accuracy": 0.9892945374691189,
      "eval_overall_f1": 0.9463171036204745,
      "eval_overall_precision": 0.9404466501240695,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.311,
      "eval_samples_per_second": 546.599,
      "eval_steps_per_second": 9.646,
      "step": 6816
    },
    {
      "epoch": 72.0,
      "grad_norm": 0.02006682939827442,
      "learning_rate": 1.4000000000000001e-05,
      "loss": 0.0035,
      "step": 6912
    },
    {
      "epoch": 72.0,
      "eval_LOCATION_f1": 0.9322033898305084,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9166666666666666,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.926984126984127,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9299363057324841,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05865481495857239,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9475,
      "eval_overall_precision": 0.9427860696517413,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3119,
      "eval_samples_per_second": 544.967,
      "eval_steps_per_second": 9.617,
      "step": 6912
    },
    {
      "epoch": 73.0,
      "grad_norm": 0.29420632123947144,
      "learning_rate": 1.3500000000000001e-05,
      "loss": 0.0041,
      "step": 7008
    },
    {
      "epoch": 73.0,
      "eval_LOCATION_f1": 0.9322033898305084,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9166666666666666,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9299363057324842,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9358974358974359,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.058048855513334274,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9486858573216519,
      "eval_overall_precision": 0.9451371571072319,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3187,
      "eval_samples_per_second": 533.464,
      "eval_steps_per_second": 9.414,
      "step": 7008
    },
    {
      "epoch": 74.0,
      "grad_norm": 0.008001981303095818,
      "learning_rate": 1.3000000000000001e-05,
      "loss": 0.0039,
      "step": 7104
    },
    {
      "epoch": 74.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9419354838709677,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9605263157894737,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.0560302734375,
      "eval_overall_accuracy": 0.9895690365083722,
      "eval_overall_f1": 0.9546599496221662,
      "eval_overall_precision": 0.9570707070707071,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3089,
      "eval_samples_per_second": 550.37,
      "eval_steps_per_second": 9.712,
      "step": 7104
    },
    {
      "epoch": 75.0,
      "grad_norm": 0.07427432388067245,
      "learning_rate": 1.25e-05,
      "loss": 0.0037,
      "step": 7200
    },
    {
      "epoch": 75.0,
      "eval_LOCATION_f1": 0.9322033898305084,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9166666666666666,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.930379746835443,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.930379746835443,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05771467462182045,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9488139825218478,
      "eval_overall_precision": 0.9429280397022333,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.31,
      "eval_samples_per_second": 548.319,
      "eval_steps_per_second": 9.676,
      "step": 7200
    },
    {
      "epoch": 76.0,
      "grad_norm": 0.05366726219654083,
      "learning_rate": 1.2e-05,
      "loss": 0.0035,
      "step": 7296
    },
    {
      "epoch": 76.0,
      "eval_LOCATION_f1": 0.9322033898305084,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9166666666666666,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.930379746835443,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.930379746835443,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.057381730526685715,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9488139825218478,
      "eval_overall_precision": 0.9429280397022333,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3098,
      "eval_samples_per_second": 548.687,
      "eval_steps_per_second": 9.683,
      "step": 7296
    },
    {
      "epoch": 77.0,
      "grad_norm": 0.4681781828403473,
      "learning_rate": 1.1500000000000002e-05,
      "loss": 0.0036,
      "step": 7392
    },
    {
      "epoch": 77.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.930379746835443,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.930379746835443,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05715038254857063,
      "eval_overall_accuracy": 0.9903925336261323,
      "eval_overall_f1": 0.951310861423221,
      "eval_overall_precision": 0.9454094292803971,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3121,
      "eval_samples_per_second": 544.702,
      "eval_steps_per_second": 9.612,
      "step": 7392
    },
    {
      "epoch": 78.0,
      "grad_norm": 0.6334789395332336,
      "learning_rate": 1.1000000000000001e-05,
      "loss": 0.0038,
      "step": 7488
    },
    {
      "epoch": 78.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.930379746835443,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.930379746835443,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.057352446019649506,
      "eval_overall_accuracy": 0.9903925336261323,
      "eval_overall_f1": 0.951310861423221,
      "eval_overall_precision": 0.9454094292803971,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.31,
      "eval_samples_per_second": 548.358,
      "eval_steps_per_second": 9.677,
      "step": 7488
    },
    {
      "epoch": 79.0,
      "grad_norm": 0.03028332069516182,
      "learning_rate": 1.05e-05,
      "loss": 0.0028,
      "step": 7584
    },
    {
      "epoch": 79.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9329073482428115,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9419354838709677,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05954783782362938,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9523809523809523,
      "eval_overall_precision": 0.95,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3114,
      "eval_samples_per_second": 545.84,
      "eval_steps_per_second": 9.632,
      "step": 7584
    },
    {
      "epoch": 80.0,
      "grad_norm": 0.0474550686776638,
      "learning_rate": 1e-05,
      "loss": 0.0035,
      "step": 7680
    },
    {
      "epoch": 80.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9389067524115756,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.954248366013072,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.06183624267578125,
      "eval_overall_accuracy": 0.9895690365083722,
      "eval_overall_f1": 0.9547738693467337,
      "eval_overall_precision": 0.9547738693467337,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3111,
      "eval_samples_per_second": 546.39,
      "eval_steps_per_second": 9.642,
      "step": 7680
    },
    {
      "epoch": 81.0,
      "grad_norm": 0.017052654176950455,
      "learning_rate": 9.5e-06,
      "loss": 0.0038,
      "step": 7776
    },
    {
      "epoch": 81.0,
      "eval_LOCATION_f1": 0.9276595744680851,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9159663865546218,
      "eval_LOCATION_recall": 0.9396551724137931,
      "eval_ORGANIZATION_f1": 0.9423076923076923,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9545454545454546,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05986158177256584,
      "eval_overall_accuracy": 0.9892945374691189,
      "eval_overall_f1": 0.9522613065326633,
      "eval_overall_precision": 0.9522613065326633,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3146,
      "eval_samples_per_second": 540.371,
      "eval_steps_per_second": 9.536,
      "step": 7776
    },
    {
      "epoch": 82.0,
      "grad_norm": 0.007322367280721664,
      "learning_rate": 9e-06,
      "loss": 0.0029,
      "step": 7872
    },
    {
      "epoch": 82.0,
      "eval_LOCATION_f1": 0.9276595744680851,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9159663865546218,
      "eval_LOCATION_recall": 0.9396551724137931,
      "eval_ORGANIZATION_f1": 0.930379746835443,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.930379746835443,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.058376602828502655,
      "eval_overall_accuracy": 0.9892945374691189,
      "eval_overall_f1": 0.9475,
      "eval_overall_precision": 0.9427860696517413,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3101,
      "eval_samples_per_second": 548.291,
      "eval_steps_per_second": 9.676,
      "step": 7872
    },
    {
      "epoch": 83.0,
      "grad_norm": 0.009456290863454342,
      "learning_rate": 8.500000000000002e-06,
      "loss": 0.0037,
      "step": 7968
    },
    {
      "epoch": 83.0,
      "eval_LOCATION_f1": 0.9276595744680851,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9159663865546218,
      "eval_LOCATION_recall": 0.9396551724137931,
      "eval_ORGANIZATION_f1": 0.9329073482428115,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9419354838709677,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.0632471889257431,
      "eval_overall_accuracy": 0.9892945374691189,
      "eval_overall_f1": 0.9485570890840652,
      "eval_overall_precision": 0.9473684210526315,
      "eval_overall_recall": 0.949748743718593,
      "eval_runtime": 0.3227,
      "eval_samples_per_second": 526.812,
      "eval_steps_per_second": 9.297,
      "step": 7968
    },
    {
      "epoch": 84.0,
      "grad_norm": 0.018801087513566017,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.003,
      "step": 8064
    },
    {
      "epoch": 84.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9329073482428115,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9419354838709677,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05650152266025543,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9510664993726474,
      "eval_overall_precision": 0.949874686716792,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3095,
      "eval_samples_per_second": 549.262,
      "eval_steps_per_second": 9.693,
      "step": 8064
    },
    {
      "epoch": 85.0,
      "grad_norm": 0.47235676646232605,
      "learning_rate": 7.5e-06,
      "loss": 0.0022,
      "step": 8160
    },
    {
      "epoch": 85.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9329073482428115,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9419354838709677,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.06099299341440201,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9510664993726474,
      "eval_overall_precision": 0.949874686716792,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3118,
      "eval_samples_per_second": 545.225,
      "eval_steps_per_second": 9.622,
      "step": 8160
    },
    {
      "epoch": 86.0,
      "grad_norm": 0.631196916103363,
      "learning_rate": 7.000000000000001e-06,
      "loss": 0.0033,
      "step": 8256
    },
    {
      "epoch": 86.0,
      "eval_LOCATION_f1": 0.9406779661016949,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.925,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9329073482428115,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9419354838709677,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.06001315265893936,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9523809523809523,
      "eval_overall_precision": 0.95,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3155,
      "eval_samples_per_second": 538.778,
      "eval_steps_per_second": 9.508,
      "step": 8256
    },
    {
      "epoch": 87.0,
      "grad_norm": 0.04047682508826256,
      "learning_rate": 6.5000000000000004e-06,
      "loss": 0.0023,
      "step": 8352
    },
    {
      "epoch": 87.0,
      "eval_LOCATION_f1": 0.9446808510638298,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9327731092436975,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9299363057324842,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9358974358974359,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.06239019334316254,
      "eval_overall_accuracy": 0.9892945374691189,
      "eval_overall_f1": 0.9523809523809523,
      "eval_overall_precision": 0.95,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3102,
      "eval_samples_per_second": 548.006,
      "eval_steps_per_second": 9.671,
      "step": 8352
    },
    {
      "epoch": 88.0,
      "grad_norm": 0.043790630996227264,
      "learning_rate": 6e-06,
      "loss": 0.002,
      "step": 8448
    },
    {
      "epoch": 88.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9389067524115756,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.954248366013072,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.06106296554207802,
      "eval_overall_accuracy": 0.9892945374691189,
      "eval_overall_f1": 0.9534591194968555,
      "eval_overall_precision": 0.9546599496221663,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3116,
      "eval_samples_per_second": 545.527,
      "eval_steps_per_second": 9.627,
      "step": 8448
    },
    {
      "epoch": 89.0,
      "grad_norm": 0.6659416556358337,
      "learning_rate": 5.500000000000001e-06,
      "loss": 0.0027,
      "step": 8544
    },
    {
      "epoch": 89.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9358974358974359,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.948051948051948,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.06123339757323265,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9522613065326633,
      "eval_overall_precision": 0.9522613065326633,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3182,
      "eval_samples_per_second": 534.197,
      "eval_steps_per_second": 9.427,
      "step": 8544
    },
    {
      "epoch": 90.0,
      "grad_norm": 0.07254405319690704,
      "learning_rate": 5e-06,
      "loss": 0.0029,
      "step": 8640
    },
    {
      "epoch": 90.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9329073482428115,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9419354838709677,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.061665747314691544,
      "eval_overall_accuracy": 0.9895690365083722,
      "eval_overall_f1": 0.9510664993726474,
      "eval_overall_precision": 0.949874686716792,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3093,
      "eval_samples_per_second": 549.619,
      "eval_steps_per_second": 9.699,
      "step": 8640
    },
    {
      "epoch": 91.0,
      "grad_norm": 0.580601155757904,
      "learning_rate": 4.5e-06,
      "loss": 0.0028,
      "step": 8736
    },
    {
      "epoch": 91.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9329073482428115,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9419354838709677,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.05955931544303894,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9510664993726474,
      "eval_overall_precision": 0.949874686716792,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3086,
      "eval_samples_per_second": 550.795,
      "eval_steps_per_second": 9.72,
      "step": 8736
    },
    {
      "epoch": 92.0,
      "grad_norm": 0.04569024220108986,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.003,
      "step": 8832
    },
    {
      "epoch": 92.0,
      "eval_LOCATION_f1": 0.9446808510638298,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9327731092436975,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9358974358974359,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.948051948051948,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.062164369970560074,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9547738693467337,
      "eval_overall_precision": 0.9547738693467337,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3104,
      "eval_samples_per_second": 547.616,
      "eval_steps_per_second": 9.664,
      "step": 8832
    },
    {
      "epoch": 93.0,
      "grad_norm": 0.031352296471595764,
      "learning_rate": 3.5000000000000004e-06,
      "loss": 0.0027,
      "step": 8928
    },
    {
      "epoch": 93.0,
      "eval_LOCATION_f1": 0.9446808510638298,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9327731092436975,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9329073482428115,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9419354838709677,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.061965953558683395,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9535759096612295,
      "eval_overall_precision": 0.9523809523809523,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3103,
      "eval_samples_per_second": 547.912,
      "eval_steps_per_second": 9.669,
      "step": 8928
    },
    {
      "epoch": 94.0,
      "grad_norm": 0.0779605507850647,
      "learning_rate": 3e-06,
      "loss": 0.0026,
      "step": 9024
    },
    {
      "epoch": 94.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9389067524115756,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.954248366013072,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.060235053300857544,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9534591194968555,
      "eval_overall_precision": 0.9546599496221663,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3114,
      "eval_samples_per_second": 545.892,
      "eval_steps_per_second": 9.633,
      "step": 9024
    },
    {
      "epoch": 95.0,
      "grad_norm": 0.010853619314730167,
      "learning_rate": 2.5e-06,
      "loss": 0.0021,
      "step": 9120
    },
    {
      "epoch": 95.0,
      "eval_LOCATION_f1": 0.9446808510638298,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9327731092436975,
      "eval_LOCATION_recall": 0.9568965517241379,
      "eval_ORGANIZATION_f1": 0.9423076923076923,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9545454545454546,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.06117314100265503,
      "eval_overall_accuracy": 0.9903925336261323,
      "eval_overall_f1": 0.957286432160804,
      "eval_overall_precision": 0.957286432160804,
      "eval_overall_recall": 0.957286432160804,
      "eval_runtime": 0.3093,
      "eval_samples_per_second": 549.565,
      "eval_steps_per_second": 9.698,
      "step": 9120
    },
    {
      "epoch": 96.0,
      "grad_norm": 0.11458369344472885,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0024,
      "step": 9216
    },
    {
      "epoch": 96.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9423076923076923,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9545454545454546,
      "eval_ORGANIZATION_recall": 0.930379746835443,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.061794668436050415,
      "eval_overall_accuracy": 0.990118034586879,
      "eval_overall_f1": 0.9547738693467337,
      "eval_overall_precision": 0.9547738693467337,
      "eval_overall_recall": 0.9547738693467337,
      "eval_runtime": 0.3159,
      "eval_samples_per_second": 538.073,
      "eval_steps_per_second": 9.495,
      "step": 9216
    },
    {
      "epoch": 97.0,
      "grad_norm": 0.01010463573038578,
      "learning_rate": 1.5e-06,
      "loss": 0.0029,
      "step": 9312
    },
    {
      "epoch": 97.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9329073482428115,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9419354838709677,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.06125541031360626,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9510664993726474,
      "eval_overall_precision": 0.949874686716792,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3105,
      "eval_samples_per_second": 547.451,
      "eval_steps_per_second": 9.661,
      "step": 9312
    },
    {
      "epoch": 98.0,
      "grad_norm": 0.13008566200733185,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0023,
      "step": 9408
    },
    {
      "epoch": 98.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9329073482428115,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9419354838709677,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.06109674647450447,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9510664993726474,
      "eval_overall_precision": 0.949874686716792,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3121,
      "eval_samples_per_second": 544.648,
      "eval_steps_per_second": 9.611,
      "step": 9408
    },
    {
      "epoch": 99.0,
      "grad_norm": 0.004328798037022352,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0025,
      "step": 9504
    },
    {
      "epoch": 99.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9329073482428115,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9419354838709677,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.061596937477588654,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9510664993726474,
      "eval_overall_precision": 0.949874686716792,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3096,
      "eval_samples_per_second": 549.121,
      "eval_steps_per_second": 9.69,
      "step": 9504
    },
    {
      "epoch": 100.0,
      "grad_norm": 0.015455431304872036,
      "learning_rate": 0.0,
      "loss": 0.0024,
      "step": 9600
    },
    {
      "epoch": 100.0,
      "eval_LOCATION_f1": 0.9361702127659575,
      "eval_LOCATION_number": 116,
      "eval_LOCATION_precision": 0.9243697478991597,
      "eval_LOCATION_recall": 0.9482758620689655,
      "eval_ORGANIZATION_f1": 0.9329073482428115,
      "eval_ORGANIZATION_number": 158,
      "eval_ORGANIZATION_precision": 0.9419354838709677,
      "eval_ORGANIZATION_recall": 0.9240506329113924,
      "eval_PERSON_f1": 0.9879518072289156,
      "eval_PERSON_number": 124,
      "eval_PERSON_precision": 0.984,
      "eval_PERSON_recall": 0.9919354838709677,
      "eval_loss": 0.06169163063168526,
      "eval_overall_accuracy": 0.9898435355476256,
      "eval_overall_f1": 0.9510664993726474,
      "eval_overall_precision": 0.949874686716792,
      "eval_overall_recall": 0.9522613065326633,
      "eval_runtime": 0.3096,
      "eval_samples_per_second": 549.148,
      "eval_steps_per_second": 9.691,
      "step": 9600
    },
    {
      "epoch": 100.0,
      "step": 9600,
      "total_flos": 4349075329087488.0,
      "train_loss": 0.030119475084356962,
      "train_runtime": 970.0979,
      "train_samples_per_second": 157.819,
      "train_steps_per_second": 9.896
    }
  ],
  "logging_steps": 500,
  "max_steps": 9600,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 100,
  "save_steps": 500,
  "total_flos": 4349075329087488.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}