trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 100.0,
  "eval_steps": 500,
  "global_step": 10600,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 1.8945987224578857,
      "learning_rate": 4.9500000000000004e-05,
      "loss": 1.1381,
      "step": 106
    },
    {
      "epoch": 1.0,
      "eval_LOCATION_f1": 0.0,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.0,
      "eval_LOCATION_recall": 0.0,
      "eval_ORGANIZATION_f1": 0.0,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.0,
      "eval_ORGANIZATION_recall": 0.0,
      "eval_PERSON_f1": 0.0,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.0,
      "eval_PERSON_recall": 0.0,
      "eval_QUANTITY_f1": 0.0,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.0,
      "eval_QUANTITY_recall": 0.0,
      "eval_TIME_f1": 0.0,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.0,
      "eval_TIME_recall": 0.0,
      "eval_loss": 0.6853168606758118,
      "eval_overall_accuracy": 0.8448836627470603,
      "eval_overall_f1": 0.0,
      "eval_overall_precision": 0.0,
      "eval_overall_recall": 0.0,
      "eval_runtime": 0.3663,
      "eval_samples_per_second": 510.496,
      "eval_steps_per_second": 8.19,
      "step": 106
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.1807020902633667,
      "learning_rate": 4.9e-05,
      "loss": 0.6775,
      "step": 212
    },
    {
      "epoch": 2.0,
      "eval_LOCATION_f1": 0.0,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.0,
      "eval_LOCATION_recall": 0.0,
      "eval_ORGANIZATION_f1": 0.0,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.0,
      "eval_ORGANIZATION_recall": 0.0,
      "eval_PERSON_f1": 0.0,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.0,
      "eval_PERSON_recall": 0.0,
      "eval_QUANTITY_f1": 0.0,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.0,
      "eval_QUANTITY_recall": 0.0,
      "eval_TIME_f1": 0.0,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.0,
      "eval_TIME_recall": 0.0,
      "eval_loss": 0.5612640976905823,
      "eval_overall_accuracy": 0.8466349762321741,
      "eval_overall_f1": 0.0,
      "eval_overall_precision": 0.0,
      "eval_overall_recall": 0.0,
      "eval_runtime": 0.3703,
      "eval_samples_per_second": 505.03,
      "eval_steps_per_second": 8.102,
      "step": 212
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.3289912939071655,
      "learning_rate": 4.85e-05,
      "loss": 0.5563,
      "step": 318
    },
    {
      "epoch": 3.0,
      "eval_LOCATION_f1": 0.1818181818181818,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.5,
      "eval_LOCATION_recall": 0.1111111111111111,
      "eval_ORGANIZATION_f1": 0.021276595744680854,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.05263157894736842,
      "eval_ORGANIZATION_recall": 0.013333333333333334,
      "eval_PERSON_f1": 0.2980392156862745,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.3392857142857143,
      "eval_PERSON_recall": 0.26573426573426573,
      "eval_QUANTITY_f1": 0.0,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.0,
      "eval_QUANTITY_recall": 0.0,
      "eval_TIME_f1": 0.18181818181818185,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.2222222222222222,
      "eval_TIME_recall": 0.15384615384615385,
      "eval_loss": 0.42658355832099915,
      "eval_overall_accuracy": 0.877658243682762,
      "eval_overall_f1": 0.201980198019802,
      "eval_overall_precision": 0.3072289156626506,
      "eval_overall_recall": 0.1504424778761062,
      "eval_runtime": 0.3682,
      "eval_samples_per_second": 507.92,
      "eval_steps_per_second": 8.148,
      "step": 318
    },
    {
      "epoch": 4.0,
      "grad_norm": 2.218233346939087,
      "learning_rate": 4.8e-05,
      "loss": 0.4194,
      "step": 424
    },
    {
      "epoch": 4.0,
      "eval_LOCATION_f1": 0.4426229508196721,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.54,
      "eval_LOCATION_recall": 0.375,
      "eval_ORGANIZATION_f1": 0.3597122302158273,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.390625,
      "eval_ORGANIZATION_recall": 0.3333333333333333,
      "eval_PERSON_f1": 0.5964912280701755,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.5125628140703518,
      "eval_PERSON_recall": 0.7132867132867133,
      "eval_QUANTITY_f1": 0.0,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.0,
      "eval_QUANTITY_recall": 0.0,
      "eval_TIME_f1": 0.5925925925925927,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.5714285714285714,
      "eval_TIME_recall": 0.6153846153846154,
      "eval_loss": 0.3157876431941986,
      "eval_overall_accuracy": 0.9126845133850388,
      "eval_overall_f1": 0.4892086330935252,
      "eval_overall_precision": 0.47752808988764045,
      "eval_overall_recall": 0.5014749262536873,
      "eval_runtime": 0.3699,
      "eval_samples_per_second": 505.61,
      "eval_steps_per_second": 8.111,
      "step": 424
    },
    {
      "epoch": 5.0,
      "grad_norm": 1.1470197439193726,
      "learning_rate": 4.75e-05,
      "loss": 0.3199,
      "step": 530
    },
    {
      "epoch": 5.0,
      "eval_LOCATION_f1": 0.625,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.5681818181818182,
      "eval_LOCATION_recall": 0.6944444444444444,
      "eval_ORGANIZATION_f1": 0.47500000000000003,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.4470588235294118,
      "eval_ORGANIZATION_recall": 0.5066666666666667,
      "eval_PERSON_f1": 0.764525993883792,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.6793478260869565,
      "eval_PERSON_recall": 0.8741258741258742,
      "eval_QUANTITY_f1": 0.35294117647058826,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.32142857142857145,
      "eval_QUANTITY_recall": 0.391304347826087,
      "eval_TIME_f1": 0.7307692307692306,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7307692307692307,
      "eval_TIME_recall": 0.7307692307692307,
      "eval_loss": 0.25322219729423523,
      "eval_overall_accuracy": 0.9301976482361771,
      "eval_overall_f1": 0.6426666666666666,
      "eval_overall_precision": 0.5863746958637469,
      "eval_overall_recall": 0.7109144542772862,
      "eval_runtime": 0.3708,
      "eval_samples_per_second": 504.27,
      "eval_steps_per_second": 8.09,
      "step": 530
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.8425076603889465,
      "learning_rate": 4.7e-05,
      "loss": 0.2663,
      "step": 636
    },
    {
      "epoch": 6.0,
      "eval_LOCATION_f1": 0.679245283018868,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6206896551724138,
      "eval_LOCATION_recall": 0.75,
      "eval_ORGANIZATION_f1": 0.5764705882352941,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5157894736842106,
      "eval_ORGANIZATION_recall": 0.6533333333333333,
      "eval_PERSON_f1": 0.8125,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7344632768361582,
      "eval_PERSON_recall": 0.9090909090909091,
      "eval_QUANTITY_f1": 0.35294117647058826,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.32142857142857145,
      "eval_QUANTITY_recall": 0.391304347826087,
      "eval_TIME_f1": 0.7241379310344829,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.65625,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.2125054895877838,
      "eval_overall_accuracy": 0.9362021516137102,
      "eval_overall_f1": 0.6939313984168866,
      "eval_overall_precision": 0.6276849642004774,
      "eval_overall_recall": 0.775811209439528,
      "eval_runtime": 0.3723,
      "eval_samples_per_second": 502.303,
      "eval_steps_per_second": 8.058,
      "step": 636
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.9470153450965881,
      "learning_rate": 4.6500000000000005e-05,
      "loss": 0.2335,
      "step": 742
    },
    {
      "epoch": 7.0,
      "eval_LOCATION_f1": 0.751592356687898,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6941176470588235,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.5862068965517241,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5151515151515151,
      "eval_ORGANIZATION_recall": 0.68,
      "eval_PERSON_f1": 0.8148148148148148,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7292817679558011,
      "eval_PERSON_recall": 0.9230769230769231,
      "eval_QUANTITY_f1": 0.4444444444444444,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.3870967741935484,
      "eval_QUANTITY_recall": 0.5217391304347826,
      "eval_TIME_f1": 0.7017543859649122,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6451612903225806,
      "eval_TIME_recall": 0.7692307692307693,
      "eval_loss": 0.1960044652223587,
      "eval_overall_accuracy": 0.9392044033024769,
      "eval_overall_f1": 0.7154046997389034,
      "eval_overall_precision": 0.6416861826697893,
      "eval_overall_recall": 0.8082595870206489,
      "eval_runtime": 0.3721,
      "eval_samples_per_second": 502.608,
      "eval_steps_per_second": 8.063,
      "step": 742
    },
    {
      "epoch": 8.0,
      "grad_norm": 1.4180433750152588,
      "learning_rate": 4.600000000000001e-05,
      "loss": 0.2093,
      "step": 848
    },
    {
      "epoch": 8.0,
      "eval_LOCATION_f1": 0.7625000000000001,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6931818181818182,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.6101694915254237,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5294117647058824,
      "eval_ORGANIZATION_recall": 0.72,
      "eval_PERSON_f1": 0.8490566037735848,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7714285714285715,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.47058823529411764,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.42857142857142855,
      "eval_QUANTITY_recall": 0.5217391304347826,
      "eval_TIME_f1": 0.7457627118644068,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6666666666666666,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.17360921204090118,
      "eval_overall_accuracy": 0.9449587190392794,
      "eval_overall_f1": 0.7424836601307189,
      "eval_overall_precision": 0.6666666666666666,
      "eval_overall_recall": 0.8377581120943953,
      "eval_runtime": 0.3739,
      "eval_samples_per_second": 500.185,
      "eval_steps_per_second": 8.024,
      "step": 848
    },
    {
      "epoch": 9.0,
      "grad_norm": 1.8917080163955688,
      "learning_rate": 4.55e-05,
      "loss": 0.1945,
      "step": 954
    },
    {
      "epoch": 9.0,
      "eval_LOCATION_f1": 0.7682926829268292,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6847826086956522,
      "eval_LOCATION_recall": 0.875,
      "eval_ORGANIZATION_f1": 0.6179775280898877,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5339805825242718,
      "eval_ORGANIZATION_recall": 0.7333333333333333,
      "eval_PERSON_f1": 0.8553054662379421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7916666666666666,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.4814814814814815,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.41935483870967744,
      "eval_QUANTITY_recall": 0.5652173913043478,
      "eval_TIME_f1": 0.7241379310344829,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.65625,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.17519919574260712,
      "eval_overall_accuracy": 0.9439579684763573,
      "eval_overall_f1": 0.7450980392156863,
      "eval_overall_precision": 0.6690140845070423,
      "eval_overall_recall": 0.8407079646017699,
      "eval_runtime": 0.3696,
      "eval_samples_per_second": 505.937,
      "eval_steps_per_second": 8.117,
      "step": 954
    },
    {
      "epoch": 10.0,
      "grad_norm": 2.473076343536377,
      "learning_rate": 4.5e-05,
      "loss": 0.1804,
      "step": 1060
    },
    {
      "epoch": 10.0,
      "eval_LOCATION_f1": 0.779874213836478,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7126436781609196,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.627906976744186,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5567010309278351,
      "eval_ORGANIZATION_recall": 0.72,
      "eval_PERSON_f1": 0.8434504792332268,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7764705882352941,
      "eval_PERSON_recall": 0.9230769230769231,
      "eval_QUANTITY_f1": 0.5423728813559322,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.4444444444444444,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.7457627118644068,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6666666666666666,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.16381332278251648,
      "eval_overall_accuracy": 0.9457092819614711,
      "eval_overall_f1": 0.7506561679790027,
      "eval_overall_precision": 0.6761229314420804,
      "eval_overall_recall": 0.8436578171091446,
      "eval_runtime": 0.3732,
      "eval_samples_per_second": 501.085,
      "eval_steps_per_second": 8.039,
      "step": 1060
    },
    {
      "epoch": 11.0,
      "grad_norm": 1.6517131328582764,
      "learning_rate": 4.4500000000000004e-05,
      "loss": 0.1692,
      "step": 1166
    },
    {
      "epoch": 11.0,
      "eval_LOCATION_f1": 0.823529411764706,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7777777777777778,
      "eval_LOCATION_recall": 0.875,
      "eval_ORGANIZATION_f1": 0.6775956284153005,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5740740740740741,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8444444444444446,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.7732558139534884,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.5357142857142857,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.45454545454545453,
      "eval_QUANTITY_recall": 0.6521739130434783,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.15897886455059052,
      "eval_overall_accuracy": 0.9499624718538904,
      "eval_overall_f1": 0.7769028871391076,
      "eval_overall_precision": 0.6997635933806147,
      "eval_overall_recall": 0.8731563421828908,
      "eval_runtime": 0.3708,
      "eval_samples_per_second": 504.355,
      "eval_steps_per_second": 8.091,
      "step": 1166
    },
    {
      "epoch": 12.0,
      "grad_norm": 2.3833324909210205,
      "learning_rate": 4.4000000000000006e-05,
      "loss": 0.1645,
      "step": 1272
    },
    {
      "epoch": 12.0,
      "eval_LOCATION_f1": 0.7947019867549668,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.759493670886076,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.6511627906976744,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5773195876288659,
      "eval_ORGANIZATION_recall": 0.7466666666666667,
      "eval_PERSON_f1": 0.8571428571428571,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8,
      "eval_PERSON_recall": 0.9230769230769231,
      "eval_QUANTITY_f1": 0.5925925925925926,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5161290322580645,
      "eval_QUANTITY_recall": 0.6956521739130435,
      "eval_TIME_f1": 0.8518518518518519,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.8214285714285714,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.1452288031578064,
      "eval_overall_accuracy": 0.9514635976982737,
      "eval_overall_f1": 0.7767253044654939,
      "eval_overall_precision": 0.7175,
      "eval_overall_recall": 0.8466076696165191,
      "eval_runtime": 0.3708,
      "eval_samples_per_second": 504.307,
      "eval_steps_per_second": 8.09,
      "step": 1272
    },
    {
      "epoch": 13.0,
      "grad_norm": 1.4393219947814941,
      "learning_rate": 4.35e-05,
      "loss": 0.1572,
      "step": 1378
    },
    {
      "epoch": 13.0,
      "eval_LOCATION_f1": 0.8079470198675496,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7721518987341772,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.6775956284153005,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5740740740740741,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.8562300319488818,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.788235294117647,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6923076923076923,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.6206896551724138,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8070175438596492,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7419354838709677,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.1519501805305481,
      "eval_overall_accuracy": 0.9514635976982737,
      "eval_overall_f1": 0.7883597883597884,
      "eval_overall_precision": 0.7146282973621103,
      "eval_overall_recall": 0.8790560471976401,
      "eval_runtime": 0.3713,
      "eval_samples_per_second": 503.618,
      "eval_steps_per_second": 8.079,
      "step": 1378
    },
    {
      "epoch": 14.0,
      "grad_norm": 1.3148809671401978,
      "learning_rate": 4.3e-05,
      "loss": 0.1538,
      "step": 1484
    },
    {
      "epoch": 14.0,
      "eval_LOCATION_f1": 0.7972972972972973,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7763157894736842,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.6630434782608695,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5596330275229358,
      "eval_ORGANIZATION_recall": 0.8133333333333334,
      "eval_PERSON_f1": 0.8664495114006516,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8109756097560976,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.7169811320754716,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.6333333333333333,
      "eval_QUANTITY_recall": 0.8260869565217391,
      "eval_TIME_f1": 0.8214285714285715,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7666666666666667,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.1491760015487671,
      "eval_overall_accuracy": 0.9509632224168126,
      "eval_overall_f1": 0.788770053475936,
      "eval_overall_precision": 0.7212713936430318,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.3732,
      "eval_samples_per_second": 501.038,
      "eval_steps_per_second": 8.038,
      "step": 1484
    },
    {
      "epoch": 15.0,
      "grad_norm": 0.9081311225891113,
      "learning_rate": 4.25e-05,
      "loss": 0.1483,
      "step": 1590
    },
    {
      "epoch": 15.0,
      "eval_LOCATION_f1": 0.7919463087248323,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7662337662337663,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.6666666666666666,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.59375,
      "eval_ORGANIZATION_recall": 0.76,
      "eval_PERSON_f1": 0.8721311475409836,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8209876543209876,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.6923076923076923,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.6206896551724138,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8518518518518519,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.8214285714285714,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.1384066790342331,
      "eval_overall_accuracy": 0.9549662246685013,
      "eval_overall_f1": 0.7934336525307797,
      "eval_overall_precision": 0.7397959183673469,
      "eval_overall_recall": 0.855457227138643,
      "eval_runtime": 0.3693,
      "eval_samples_per_second": 506.326,
      "eval_steps_per_second": 8.123,
      "step": 1590
    },
    {
      "epoch": 16.0,
      "grad_norm": 1.228157877922058,
      "learning_rate": 4.2e-05,
      "loss": 0.145,
      "step": 1696
    },
    {
      "epoch": 16.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.6737967914438503,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5625,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8681672025723473,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8035714285714286,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.7407407407407407,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.6451612903225806,
      "eval_QUANTITY_recall": 0.8695652173913043,
      "eval_TIME_f1": 0.7857142857142856,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7333333333333333,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.15306319296360016,
      "eval_overall_accuracy": 0.9497122842131599,
      "eval_overall_f1": 0.7874015748031495,
      "eval_overall_precision": 0.7092198581560284,
      "eval_overall_recall": 0.8849557522123894,
      "eval_runtime": 0.3725,
      "eval_samples_per_second": 502.005,
      "eval_steps_per_second": 8.054,
      "step": 1696
    },
    {
      "epoch": 17.0,
      "grad_norm": 0.6019057035446167,
      "learning_rate": 4.15e-05,
      "loss": 0.1378,
      "step": 1802
    },
    {
      "epoch": 17.0,
      "eval_LOCATION_f1": 0.7770700636942675,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7176470588235294,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.6703296703296704,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5700934579439252,
      "eval_ORGANIZATION_recall": 0.8133333333333334,
      "eval_PERSON_f1": 0.8778877887788779,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.83125,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.7272727272727273,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.625,
      "eval_QUANTITY_recall": 0.8695652173913043,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.146218940615654,
      "eval_overall_accuracy": 0.9514635976982737,
      "eval_overall_f1": 0.7925531914893617,
      "eval_overall_precision": 0.7215496368038741,
      "eval_overall_recall": 0.8790560471976401,
      "eval_runtime": 0.3673,
      "eval_samples_per_second": 509.115,
      "eval_steps_per_second": 8.168,
      "step": 1802
    },
    {
      "epoch": 18.0,
      "grad_norm": 0.6048269867897034,
      "learning_rate": 4.1e-05,
      "loss": 0.137,
      "step": 1908
    },
    {
      "epoch": 18.0,
      "eval_LOCATION_f1": 0.7898089171974522,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7294117647058823,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.6486486486486486,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5454545454545454,
      "eval_ORGANIZATION_recall": 0.8,
      "eval_PERSON_f1": 0.8837209302325583,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8417721518987342,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.7407407407407407,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.6451612903225806,
      "eval_QUANTITY_recall": 0.8695652173913043,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.143734872341156,
      "eval_overall_accuracy": 0.9522141606204654,
      "eval_overall_f1": 0.7925531914893617,
      "eval_overall_precision": 0.7215496368038741,
      "eval_overall_recall": 0.8790560471976401,
      "eval_runtime": 0.3712,
      "eval_samples_per_second": 503.734,
      "eval_steps_per_second": 8.081,
      "step": 1908
    },
    {
      "epoch": 19.0,
      "grad_norm": 0.9862064719200134,
      "learning_rate": 4.05e-05,
      "loss": 0.1316,
      "step": 2014
    },
    {
      "epoch": 19.0,
      "eval_LOCATION_f1": 0.7792207792207793,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7317073170731707,
      "eval_LOCATION_recall": 0.8333333333333334,
      "eval_ORGANIZATION_f1": 0.6931818181818181,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6039603960396039,
      "eval_ORGANIZATION_recall": 0.8133333333333334,
      "eval_PERSON_f1": 0.868421052631579,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8198757763975155,
      "eval_PERSON_recall": 0.9230769230769231,
      "eval_QUANTITY_f1": 0.7407407407407407,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.6451612903225806,
      "eval_QUANTITY_recall": 0.8695652173913043,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13715997338294983,
      "eval_overall_accuracy": 0.9542156617463098,
      "eval_overall_f1": 0.7940780619111709,
      "eval_overall_precision": 0.7301980198019802,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.3712,
      "eval_samples_per_second": 503.829,
      "eval_steps_per_second": 8.083,
      "step": 2014
    },
    {
      "epoch": 20.0,
      "grad_norm": 1.5772024393081665,
      "learning_rate": 4e-05,
      "loss": 0.1274,
      "step": 2120
    },
    {
      "epoch": 20.0,
      "eval_LOCATION_f1": 0.7875000000000001,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7159090909090909,
      "eval_LOCATION_recall": 0.875,
      "eval_ORGANIZATION_f1": 0.6918918918918919,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5818181818181818,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8721311475409836,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8209876543209876,
      "eval_PERSON_recall": 0.9300699300699301,
      "eval_QUANTITY_f1": 0.7272727272727273,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.625,
      "eval_QUANTITY_recall": 0.8695652173913043,
      "eval_TIME_f1": 0.7586206896551724,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6875,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1509571373462677,
      "eval_overall_accuracy": 0.950212659494621,
      "eval_overall_f1": 0.7916120576671036,
      "eval_overall_precision": 0.7122641509433962,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.3724,
      "eval_samples_per_second": 502.213,
      "eval_steps_per_second": 8.057,
      "step": 2120
    },
    {
      "epoch": 21.0,
      "grad_norm": 1.3061535358428955,
      "learning_rate": 3.9500000000000005e-05,
      "loss": 0.1279,
      "step": 2226
    },
    {
      "epoch": 21.0,
      "eval_LOCATION_f1": 0.7898089171974522,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7294117647058823,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7111111111111111,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6095238095238096,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8737864077669902,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8132530120481928,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.7407407407407407,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.6451612903225806,
      "eval_QUANTITY_recall": 0.8695652173913043,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.14436930418014526,
      "eval_overall_accuracy": 0.9532149111833875,
      "eval_overall_f1": 0.8052980132450331,
      "eval_overall_precision": 0.7307692307692307,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.3764,
      "eval_samples_per_second": 496.865,
      "eval_steps_per_second": 7.971,
      "step": 2226
    },
    {
      "epoch": 22.0,
      "grad_norm": 0.9500447511672974,
      "learning_rate": 3.9000000000000006e-05,
      "loss": 0.1256,
      "step": 2332
    },
    {
      "epoch": 22.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7159090909090909,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6237623762376238,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8852459016393444,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8333333333333334,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.7142857142857143,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.6060606060606061,
      "eval_QUANTITY_recall": 0.8695652173913043,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.13468235731124878,
      "eval_overall_accuracy": 0.9567175381536153,
      "eval_overall_f1": 0.8123324396782842,
      "eval_overall_precision": 0.7444717444717445,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3693,
      "eval_samples_per_second": 506.364,
      "eval_steps_per_second": 8.123,
      "step": 2332
    },
    {
      "epoch": 23.0,
      "grad_norm": 1.3265873193740845,
      "learning_rate": 3.85e-05,
      "loss": 0.123,
      "step": 2438
    },
    {
      "epoch": 23.0,
      "eval_LOCATION_f1": 0.781456953642384,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7468354430379747,
      "eval_LOCATION_recall": 0.8194444444444444,
      "eval_ORGANIZATION_f1": 0.7262569832402236,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.625,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8831168831168831,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8242424242424242,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6785714285714286,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5757575757575758,
      "eval_QUANTITY_recall": 0.8260869565217391,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13530714809894562,
      "eval_overall_accuracy": 0.955716787590693,
      "eval_overall_f1": 0.8037383177570093,
      "eval_overall_precision": 0.7341463414634146,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3693,
      "eval_samples_per_second": 506.356,
      "eval_steps_per_second": 8.123,
      "step": 2438
    },
    {
      "epoch": 24.0,
      "grad_norm": 1.4870867729187012,
      "learning_rate": 3.8e-05,
      "loss": 0.1219,
      "step": 2544
    },
    {
      "epoch": 24.0,
      "eval_LOCATION_f1": 0.7654320987654321,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6888888888888889,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7032967032967032,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.5981308411214953,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8794788273615636,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.823170731707317,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.7017543859649124,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5882352941176471,
      "eval_QUANTITY_recall": 0.8695652173913043,
      "eval_TIME_f1": 0.7586206896551724,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6875,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.14730946719646454,
      "eval_overall_accuracy": 0.9519639729797348,
      "eval_overall_f1": 0.7911227154046998,
      "eval_overall_precision": 0.7096018735362998,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3722,
      "eval_samples_per_second": 502.446,
      "eval_steps_per_second": 8.061,
      "step": 2544
    },
    {
      "epoch": 25.0,
      "grad_norm": 1.858078956604004,
      "learning_rate": 3.7500000000000003e-05,
      "loss": 0.1193,
      "step": 2650
    },
    {
      "epoch": 25.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.735632183908046,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6464646464646465,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8896103896103895,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8303030303030303,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6785714285714286,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5757575757575758,
      "eval_QUANTITY_recall": 0.8260869565217391,
      "eval_TIME_f1": 0.7857142857142856,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7333333333333333,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1400519758462906,
      "eval_overall_accuracy": 0.9567175381536153,
      "eval_overall_f1": 0.8128342245989304,
      "eval_overall_precision": 0.7432762836185819,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.368,
      "eval_samples_per_second": 508.1,
      "eval_steps_per_second": 8.151,
      "step": 2650
    },
    {
      "epoch": 26.0,
      "grad_norm": 0.955053985118866,
      "learning_rate": 3.7e-05,
      "loss": 0.1168,
      "step": 2756
    },
    {
      "epoch": 26.0,
      "eval_LOCATION_f1": 0.7948717948717949,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7380952380952381,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7159090909090909,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6237623762376238,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8385093167701864,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6792452830188679,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.6,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7857142857142856,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7333333333333333,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13576149940490723,
      "eval_overall_accuracy": 0.955716787590693,
      "eval_overall_f1": 0.8053691275167786,
      "eval_overall_precision": 0.7389162561576355,
      "eval_overall_recall": 0.8849557522123894,
      "eval_runtime": 0.3716,
      "eval_samples_per_second": 503.212,
      "eval_steps_per_second": 8.073,
      "step": 2756
    },
    {
      "epoch": 27.0,
      "grad_norm": 1.6411521434783936,
      "learning_rate": 3.65e-05,
      "loss": 0.1132,
      "step": 2862
    },
    {
      "epoch": 27.0,
      "eval_LOCATION_f1": 0.7625000000000001,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6931818181818182,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7073170731707319,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.651685393258427,
      "eval_ORGANIZATION_recall": 0.7733333333333333,
      "eval_PERSON_f1": 0.8852459016393444,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8333333333333334,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6785714285714286,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5757575757575758,
      "eval_QUANTITY_recall": 0.8260869565217391,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.14072290062904358,
      "eval_overall_accuracy": 0.9547160370277709,
      "eval_overall_f1": 0.7972972972972973,
      "eval_overall_precision": 0.7356608478802993,
      "eval_overall_recall": 0.8702064896755162,
      "eval_runtime": 0.3708,
      "eval_samples_per_second": 504.296,
      "eval_steps_per_second": 8.09,
      "step": 2862
    },
    {
      "epoch": 28.0,
      "grad_norm": 1.1676483154296875,
      "learning_rate": 3.6e-05,
      "loss": 0.115,
      "step": 2968
    },
    {
      "epoch": 28.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7231638418079096,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6274509803921569,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8322981366459627,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6909090909090908,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.59375,
      "eval_QUANTITY_recall": 0.8260869565217391,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.13657206296920776,
      "eval_overall_accuracy": 0.9567175381536153,
      "eval_overall_f1": 0.8080536912751678,
      "eval_overall_precision": 0.7413793103448276,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3699,
      "eval_samples_per_second": 505.486,
      "eval_steps_per_second": 8.109,
      "step": 2968
    },
    {
      "epoch": 29.0,
      "grad_norm": 1.2002038955688477,
      "learning_rate": 3.55e-05,
      "loss": 0.112,
      "step": 3074
    },
    {
      "epoch": 29.0,
      "eval_LOCATION_f1": 0.8,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7469879518072289,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7386363636363636,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6435643564356436,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.9006622516556291,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8553459119496856,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6909090909090908,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.59375,
      "eval_QUANTITY_recall": 0.8260869565217391,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13843736052513123,
      "eval_overall_accuracy": 0.9559669752314236,
      "eval_overall_f1": 0.8183041722745626,
      "eval_overall_precision": 0.7524752475247525,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.371,
      "eval_samples_per_second": 504.044,
      "eval_steps_per_second": 8.086,
      "step": 3074
    },
    {
      "epoch": 30.0,
      "grad_norm": 1.0138002634048462,
      "learning_rate": 3.5e-05,
      "loss": 0.1112,
      "step": 3180
    },
    {
      "epoch": 30.0,
      "eval_LOCATION_f1": 0.7701863354037267,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.6966292134831461,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7303370786516854,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6310679611650486,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8947368421052632,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84472049689441,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13617390394210815,
      "eval_overall_accuracy": 0.955716787590693,
      "eval_overall_f1": 0.804780876494024,
      "eval_overall_precision": 0.7318840579710145,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3711,
      "eval_samples_per_second": 503.907,
      "eval_steps_per_second": 8.084,
      "step": 3180
    },
    {
      "epoch": 31.0,
      "grad_norm": 0.6963276863098145,
      "learning_rate": 3.45e-05,
      "loss": 0.1093,
      "step": 3286
    },
    {
      "epoch": 31.0,
      "eval_LOCATION_f1": 0.782051282051282,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7261904761904762,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7455621301775148,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6702127659574468,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8947368421052632,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84472049689441,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6909090909090908,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.59375,
      "eval_QUANTITY_recall": 0.8260869565217391,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.13229180872440338,
      "eval_overall_accuracy": 0.9577182887165374,
      "eval_overall_f1": 0.817320703653586,
      "eval_overall_precision": 0.755,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.369,
      "eval_samples_per_second": 506.747,
      "eval_steps_per_second": 8.13,
      "step": 3286
    },
    {
      "epoch": 32.0,
      "grad_norm": 2.4005796909332275,
      "learning_rate": 3.4000000000000007e-05,
      "loss": 0.1075,
      "step": 3392
    },
    {
      "epoch": 32.0,
      "eval_LOCATION_f1": 0.8157894736842106,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.775,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7425149700598803,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6739130434782609,
      "eval_ORGANIZATION_recall": 0.8266666666666667,
      "eval_PERSON_f1": 0.9,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8598726114649682,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6909090909090908,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.59375,
      "eval_QUANTITY_recall": 0.8260869565217391,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.12926289439201355,
      "eval_overall_accuracy": 0.9597197898423818,
      "eval_overall_f1": 0.8257887517146778,
      "eval_overall_precision": 0.7717948717948718,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3724,
      "eval_samples_per_second": 502.175,
      "eval_steps_per_second": 8.056,
      "step": 3392
    },
    {
      "epoch": 33.0,
      "grad_norm": 1.497031807899475,
      "learning_rate": 3.35e-05,
      "loss": 0.1046,
      "step": 3498
    },
    {
      "epoch": 33.0,
      "eval_LOCATION_f1": 0.7770700636942675,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7176470588235294,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.744186046511628,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6597938144329897,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.9158249158249158,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8831168831168831,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8214285714285715,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7666666666666667,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.13126622140407562,
      "eval_overall_accuracy": 0.9594696022016512,
      "eval_overall_f1": 0.819538670284939,
      "eval_overall_precision": 0.7587939698492462,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.3762,
      "eval_samples_per_second": 497.031,
      "eval_steps_per_second": 7.974,
      "step": 3498
    },
    {
      "epoch": 34.0,
      "grad_norm": 0.8285422921180725,
      "learning_rate": 3.3e-05,
      "loss": 0.1017,
      "step": 3604
    },
    {
      "epoch": 34.0,
      "eval_LOCATION_f1": 0.7848101265822784,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7209302325581395,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7314285714285714,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.64,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8823529411764706,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8282208588957055,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7586206896551724,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6875,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1426933854818344,
      "eval_overall_accuracy": 0.9547160370277709,
      "eval_overall_f1": 0.799468791500664,
      "eval_overall_precision": 0.7270531400966184,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3722,
      "eval_samples_per_second": 502.449,
      "eval_steps_per_second": 8.061,
      "step": 3604
    },
    {
      "epoch": 35.0,
      "grad_norm": 1.9264658689498901,
      "learning_rate": 3.2500000000000004e-05,
      "loss": 0.1071,
      "step": 3710
    },
    {
      "epoch": 35.0,
      "eval_LOCATION_f1": 0.779874213836478,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7126436781609196,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7231638418079096,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6274509803921569,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.9066666666666667,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8662420382165605,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.14034150540828705,
      "eval_overall_accuracy": 0.9562171628721541,
      "eval_overall_f1": 0.8112449799196788,
      "eval_overall_precision": 0.7426470588235294,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3705,
      "eval_samples_per_second": 504.772,
      "eval_steps_per_second": 8.098,
      "step": 3710
    },
    {
      "epoch": 36.0,
      "grad_norm": 1.7772150039672852,
      "learning_rate": 3.2000000000000005e-05,
      "loss": 0.1026,
      "step": 3816
    },
    {
      "epoch": 36.0,
      "eval_LOCATION_f1": 0.782051282051282,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7261904761904762,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7407407407407408,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6896551724137931,
      "eval_ORGANIZATION_recall": 0.8,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.12969908118247986,
      "eval_overall_accuracy": 0.9589692269201902,
      "eval_overall_f1": 0.8153214774281805,
      "eval_overall_precision": 0.7602040816326531,
      "eval_overall_recall": 0.8790560471976401,
      "eval_runtime": 0.3711,
      "eval_samples_per_second": 503.942,
      "eval_steps_per_second": 8.085,
      "step": 3816
    },
    {
      "epoch": 37.0,
      "grad_norm": 5.3707380294799805,
      "learning_rate": 3.15e-05,
      "loss": 0.1005,
      "step": 3922
    },
    {
      "epoch": 37.0,
      "eval_LOCATION_f1": 0.7848101265822784,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7209302325581395,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7570621468926553,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6568627450980392,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.8954248366013072,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8404907975460123,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8214285714285715,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7666666666666667,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.14334312081336975,
      "eval_overall_accuracy": 0.9549662246685013,
      "eval_overall_f1": 0.8127490039840637,
      "eval_overall_precision": 0.7391304347826086,
      "eval_overall_recall": 0.9026548672566371,
      "eval_runtime": 0.3705,
      "eval_samples_per_second": 504.68,
      "eval_steps_per_second": 8.096,
      "step": 3922
    },
    {
      "epoch": 38.0,
      "grad_norm": 1.382991075515747,
      "learning_rate": 3.1e-05,
      "loss": 0.0981,
      "step": 4028
    },
    {
      "epoch": 38.0,
      "eval_LOCATION_f1": 0.8157894736842106,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.775,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7386363636363636,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6435643564356436,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8990228013029316,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8414634146341463,
      "eval_PERSON_recall": 0.965034965034965,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8214285714285715,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7666666666666667,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.14039599895477295,
      "eval_overall_accuracy": 0.9572179134350763,
      "eval_overall_f1": 0.8192771084337348,
      "eval_overall_precision": 0.75,
      "eval_overall_recall": 0.9026548672566371,
      "eval_runtime": 0.371,
      "eval_samples_per_second": 503.978,
      "eval_steps_per_second": 8.085,
      "step": 4028
    },
    {
      "epoch": 39.0,
      "grad_norm": 0.9052088260650635,
      "learning_rate": 3.05e-05,
      "loss": 0.0982,
      "step": 4134
    },
    {
      "epoch": 39.0,
      "eval_LOCATION_f1": 0.8104575163398693,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7654320987654321,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7745664739884393,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6836734693877551,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.8888888888888888,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8343558282208589,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8214285714285715,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7666666666666667,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.1408725529909134,
      "eval_overall_accuracy": 0.957968476357268,
      "eval_overall_f1": 0.8225806451612903,
      "eval_overall_precision": 0.7555555555555555,
      "eval_overall_recall": 0.9026548672566371,
      "eval_runtime": 0.3728,
      "eval_samples_per_second": 501.569,
      "eval_steps_per_second": 8.047,
      "step": 4134
    },
    {
      "epoch": 40.0,
      "grad_norm": 1.2666678428649902,
      "learning_rate": 3e-05,
      "loss": 0.0986,
      "step": 4240
    },
    {
      "epoch": 40.0,
      "eval_LOCATION_f1": 0.7973856209150327,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7530864197530864,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7810650887573964,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7021276595744681,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8859934853420195,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8292682926829268,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6296296296296297,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5483870967741935,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.13242048025131226,
      "eval_overall_accuracy": 0.9587190392794596,
      "eval_overall_f1": 0.8211382113821138,
      "eval_overall_precision": 0.7593984962406015,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3704,
      "eval_samples_per_second": 504.864,
      "eval_steps_per_second": 8.099,
      "step": 4240
    },
    {
      "epoch": 41.0,
      "grad_norm": 1.3640594482421875,
      "learning_rate": 2.95e-05,
      "loss": 0.0963,
      "step": 4346
    },
    {
      "epoch": 41.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7674418604651163,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6804123711340206,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8888888888888888,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8343558282208589,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6538461538461539,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5862068965517241,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.1359548419713974,
      "eval_overall_accuracy": 0.9577182887165374,
      "eval_overall_f1": 0.8227334235453315,
      "eval_overall_precision": 0.76,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.3704,
      "eval_samples_per_second": 504.827,
      "eval_steps_per_second": 8.099,
      "step": 4346
    },
    {
      "epoch": 42.0,
      "grad_norm": 1.9110503196716309,
      "learning_rate": 2.9e-05,
      "loss": 0.096,
      "step": 4452
    },
    {
      "epoch": 42.0,
      "eval_LOCATION_f1": 0.8,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7469879518072289,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7719298245614036,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6875,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8983606557377047,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.845679012345679,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8070175438596492,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7419354838709677,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.1379159390926361,
      "eval_overall_accuracy": 0.9584688516387291,
      "eval_overall_f1": 0.8236877523553163,
      "eval_overall_precision": 0.7574257425742574,
      "eval_overall_recall": 0.9026548672566371,
      "eval_runtime": 0.3712,
      "eval_samples_per_second": 503.833,
      "eval_steps_per_second": 8.083,
      "step": 4452
    },
    {
      "epoch": 43.0,
      "grad_norm": 0.9015399813652039,
      "learning_rate": 2.8499999999999998e-05,
      "loss": 0.0944,
      "step": 4558
    },
    {
      "epoch": 43.0,
      "eval_LOCATION_f1": 0.8026315789473685,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7625,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.8,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7333333333333333,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.9066666666666667,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8662420382165605,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6538461538461539,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5862068965517241,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.12910011410713196,
      "eval_overall_accuracy": 0.9617212909682261,
      "eval_overall_f1": 0.8342541436464089,
      "eval_overall_precision": 0.7844155844155845,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.3719,
      "eval_samples_per_second": 502.787,
      "eval_steps_per_second": 8.066,
      "step": 4558
    },
    {
      "epoch": 44.0,
      "grad_norm": 1.5744454860687256,
      "learning_rate": 2.8000000000000003e-05,
      "loss": 0.0959,
      "step": 4664
    },
    {
      "epoch": 44.0,
      "eval_LOCATION_f1": 0.8129032258064516,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7590361445783133,
      "eval_LOCATION_recall": 0.875,
      "eval_ORGANIZATION_f1": 0.788235294117647,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7052631578947368,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.9013157894736843,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8509316770186336,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6296296296296297,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5483870967741935,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7272727272727274,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6896551724137931,
      "eval_TIME_recall": 0.7692307692307693,
      "eval_loss": 0.13805752992630005,
      "eval_overall_accuracy": 0.9569677257943457,
      "eval_overall_f1": 0.8238482384823848,
      "eval_overall_precision": 0.7619047619047619,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.3727,
      "eval_samples_per_second": 501.769,
      "eval_steps_per_second": 8.05,
      "step": 4664
    },
    {
      "epoch": 45.0,
      "grad_norm": 1.3224737644195557,
      "learning_rate": 2.7500000000000004e-05,
      "loss": 0.0935,
      "step": 4770
    },
    {
      "epoch": 45.0,
      "eval_LOCATION_f1": 0.823529411764706,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7777777777777778,
      "eval_LOCATION_recall": 0.875,
      "eval_ORGANIZATION_f1": 0.7647058823529413,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6842105263157895,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.9078947368421052,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8571428571428571,
      "eval_PERSON_recall": 0.965034965034965,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13827811181545258,
      "eval_overall_accuracy": 0.9584688516387291,
      "eval_overall_f1": 0.8249660786974219,
      "eval_overall_precision": 0.7638190954773869,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.3741,
      "eval_samples_per_second": 499.811,
      "eval_steps_per_second": 8.018,
      "step": 4770
    },
    {
      "epoch": 46.0,
      "grad_norm": 0.8794982433319092,
      "learning_rate": 2.7000000000000002e-05,
      "loss": 0.0903,
      "step": 4876
    },
    {
      "epoch": 46.0,
      "eval_LOCATION_f1": 0.8,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7469879518072289,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7514450867052023,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6632653061224489,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.9042904290429041,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85625,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6415094339622641,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5666666666666667,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8363636363636363,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7931034482758621,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.13555409014225006,
      "eval_overall_accuracy": 0.9582186639979985,
      "eval_overall_f1": 0.8227334235453315,
      "eval_overall_precision": 0.76,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.3704,
      "eval_samples_per_second": 504.861,
      "eval_steps_per_second": 8.099,
      "step": 4876
    },
    {
      "epoch": 47.0,
      "grad_norm": 1.307417869567871,
      "learning_rate": 2.6500000000000004e-05,
      "loss": 0.0933,
      "step": 4982
    },
    {
      "epoch": 47.0,
      "eval_LOCATION_f1": 0.7973856209150327,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7530864197530864,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7816091954022988,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6868686868686869,
      "eval_ORGANIZATION_recall": 0.9066666666666666,
      "eval_PERSON_f1": 0.9019607843137255,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8466257668711656,
      "eval_PERSON_recall": 0.965034965034965,
      "eval_QUANTITY_f1": 0.5964912280701754,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7272727272727274,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6896551724137931,
      "eval_TIME_recall": 0.7692307692307693,
      "eval_loss": 0.15163101255893707,
      "eval_overall_accuracy": 0.9552164123092319,
      "eval_overall_f1": 0.8161073825503355,
      "eval_overall_precision": 0.7487684729064039,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.3701,
      "eval_samples_per_second": 505.224,
      "eval_steps_per_second": 8.105,
      "step": 4982
    },
    {
      "epoch": 48.0,
      "grad_norm": 1.3334672451019287,
      "learning_rate": 2.6000000000000002e-05,
      "loss": 0.093,
      "step": 5088
    },
    {
      "epoch": 48.0,
      "eval_LOCATION_f1": 0.8,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7469879518072289,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7647058823529413,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6842105263157895,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6296296296296297,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5483870967741935,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13479116559028625,
      "eval_overall_accuracy": 0.957968476357268,
      "eval_overall_f1": 0.8168249660786975,
      "eval_overall_precision": 0.7562814070351759,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3689,
      "eval_samples_per_second": 506.907,
      "eval_steps_per_second": 8.132,
      "step": 5088
    },
    {
      "epoch": 49.0,
      "grad_norm": 2.846377372741699,
      "learning_rate": 2.5500000000000003e-05,
      "loss": 0.0906,
      "step": 5194
    },
    {
      "epoch": 49.0,
      "eval_LOCATION_f1": 0.8,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7469879518072289,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7640449438202248,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6601941747572816,
      "eval_ORGANIZATION_recall": 0.9066666666666666,
      "eval_PERSON_f1": 0.9042904290429041,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85625,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6666666666666667,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5806451612903226,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14537392556667328,
      "eval_overall_accuracy": 0.9569677257943457,
      "eval_overall_f1": 0.8214765100671141,
      "eval_overall_precision": 0.7536945812807881,
      "eval_overall_recall": 0.9026548672566371,
      "eval_runtime": 0.3727,
      "eval_samples_per_second": 501.678,
      "eval_steps_per_second": 8.048,
      "step": 5194
    },
    {
      "epoch": 50.0,
      "grad_norm": 1.0936921834945679,
      "learning_rate": 2.5e-05,
      "loss": 0.0894,
      "step": 5300
    },
    {
      "epoch": 50.0,
      "eval_LOCATION_f1": 0.8026315789473685,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7625,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7730061349693251,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7159090909090909,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8918032786885245,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8395061728395061,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6666666666666667,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5806451612903226,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8518518518518519,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.8214285714285714,
      "eval_TIME_recall": 0.8846153846153846,
      "eval_loss": 0.13398553431034088,
      "eval_overall_accuracy": 0.9609707280460346,
      "eval_overall_f1": 0.8269230769230769,
      "eval_overall_precision": 0.7737789203084833,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3704,
      "eval_samples_per_second": 504.87,
      "eval_steps_per_second": 8.1,
      "step": 5300
    },
    {
      "epoch": 51.0,
      "grad_norm": 1.3817942142486572,
      "learning_rate": 2.45e-05,
      "loss": 0.0882,
      "step": 5406
    },
    {
      "epoch": 51.0,
      "eval_LOCATION_f1": 0.8104575163398693,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7654320987654321,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7602339181286549,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6770833333333334,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.9013157894736843,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8509316770186336,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6415094339622641,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5666666666666667,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1364634782075882,
      "eval_overall_accuracy": 0.9594696022016512,
      "eval_overall_f1": 0.8233695652173914,
      "eval_overall_precision": 0.7632241813602015,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3697,
      "eval_samples_per_second": 505.839,
      "eval_steps_per_second": 8.115,
      "step": 5406
    },
    {
      "epoch": 52.0,
      "grad_norm": 0.8930407762527466,
      "learning_rate": 2.4e-05,
      "loss": 0.0886,
      "step": 5512
    },
    {
      "epoch": 52.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7701149425287357,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6767676767676768,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.9049180327868852,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8518518518518519,
      "eval_PERSON_recall": 0.965034965034965,
      "eval_QUANTITY_f1": 0.6296296296296297,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5483870967741935,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14068011939525604,
      "eval_overall_accuracy": 0.9587190392794596,
      "eval_overall_f1": 0.8221024258760108,
      "eval_overall_precision": 0.7568238213399504,
      "eval_overall_recall": 0.8997050147492626,
      "eval_runtime": 0.3699,
      "eval_samples_per_second": 505.493,
      "eval_steps_per_second": 8.11,
      "step": 5512
    },
    {
      "epoch": 53.0,
      "grad_norm": 1.4923311471939087,
      "learning_rate": 2.35e-05,
      "loss": 0.086,
      "step": 5618
    },
    {
      "epoch": 53.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7790697674418604,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6907216494845361,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.8947368421052632,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84472049689441,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6666666666666667,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5806451612903226,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7857142857142856,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7333333333333333,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13908393681049347,
      "eval_overall_accuracy": 0.9592194145609206,
      "eval_overall_f1": 0.8243243243243245,
      "eval_overall_precision": 0.7605985037406484,
      "eval_overall_recall": 0.8997050147492626,
      "eval_runtime": 0.3682,
      "eval_samples_per_second": 507.889,
      "eval_steps_per_second": 8.148,
      "step": 5618
    },
    {
      "epoch": 54.0,
      "grad_norm": 2.1307640075683594,
      "learning_rate": 2.3000000000000003e-05,
      "loss": 0.0872,
      "step": 5724
    },
    {
      "epoch": 54.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7810650887573964,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7021276595744681,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.903654485049834,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8607594936708861,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1391591876745224,
      "eval_overall_accuracy": 0.9589692269201902,
      "eval_overall_f1": 0.8256130790190735,
      "eval_overall_precision": 0.7670886075949367,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3712,
      "eval_samples_per_second": 503.825,
      "eval_steps_per_second": 8.083,
      "step": 5724
    },
    {
      "epoch": 55.0,
      "grad_norm": 2.045675277709961,
      "learning_rate": 2.25e-05,
      "loss": 0.0897,
      "step": 5830
    },
    {
      "epoch": 55.0,
      "eval_LOCATION_f1": 0.7948717948717949,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7380952380952381,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7640449438202248,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6601941747572816,
      "eval_ORGANIZATION_recall": 0.9066666666666666,
      "eval_PERSON_f1": 0.9078947368421052,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8571428571428571,
      "eval_PERSON_recall": 0.965034965034965,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7142857142857142,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6666666666666666,
      "eval_TIME_recall": 0.7692307692307693,
      "eval_loss": 0.1569395512342453,
      "eval_overall_accuracy": 0.9534650988241181,
      "eval_overall_f1": 0.8133333333333332,
      "eval_overall_precision": 0.7420924574209246,
      "eval_overall_recall": 0.8997050147492626,
      "eval_runtime": 0.3688,
      "eval_samples_per_second": 507.118,
      "eval_steps_per_second": 8.136,
      "step": 5830
    },
    {
      "epoch": 56.0,
      "grad_norm": 0.979404091835022,
      "learning_rate": 2.2000000000000003e-05,
      "loss": 0.0881,
      "step": 5936
    },
    {
      "epoch": 56.0,
      "eval_LOCATION_f1": 0.7973856209150327,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7530864197530864,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7831325301204819,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7142857142857143,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.9006622516556291,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8553459119496856,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13650928437709808,
      "eval_overall_accuracy": 0.9597197898423818,
      "eval_overall_f1": 0.8207934336525308,
      "eval_overall_precision": 0.7653061224489796,
      "eval_overall_recall": 0.8849557522123894,
      "eval_runtime": 0.372,
      "eval_samples_per_second": 502.676,
      "eval_steps_per_second": 8.064,
      "step": 5936
    },
    {
      "epoch": 57.0,
      "grad_norm": 1.4905277490615845,
      "learning_rate": 2.15e-05,
      "loss": 0.0857,
      "step": 6042
    },
    {
      "epoch": 57.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.8,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7333333333333333,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6296296296296297,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5483870967741935,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7777777777777779,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.75,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.13628903031349182,
      "eval_overall_accuracy": 0.961220915686765,
      "eval_overall_f1": 0.8273972602739725,
      "eval_overall_precision": 0.7723785166240409,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.3704,
      "eval_samples_per_second": 504.837,
      "eval_steps_per_second": 8.099,
      "step": 6042
    },
    {
      "epoch": 58.0,
      "grad_norm": 0.8609861135482788,
      "learning_rate": 2.1e-05,
      "loss": 0.0846,
      "step": 6148
    },
    {
      "epoch": 58.0,
      "eval_LOCATION_f1": 0.7973856209150327,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7530864197530864,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.783625730994152,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6979166666666666,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6296296296296297,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5483870967741935,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.14013363420963287,
      "eval_overall_accuracy": 0.9594696022016512,
      "eval_overall_f1": 0.8233695652173914,
      "eval_overall_precision": 0.7632241813602015,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3725,
      "eval_samples_per_second": 502.031,
      "eval_steps_per_second": 8.054,
      "step": 6148
    },
    {
      "epoch": 59.0,
      "grad_norm": 1.2671494483947754,
      "learning_rate": 2.05e-05,
      "loss": 0.0843,
      "step": 6254
    },
    {
      "epoch": 59.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7928994082840236,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7127659574468085,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.9078947368421052,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8571428571428571,
      "eval_PERSON_recall": 0.965034965034965,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.75,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14256823062896729,
      "eval_overall_accuracy": 0.9594696022016512,
      "eval_overall_f1": 0.8238482384823848,
      "eval_overall_precision": 0.7619047619047619,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.3732,
      "eval_samples_per_second": 501.11,
      "eval_steps_per_second": 8.039,
      "step": 6254
    },
    {
      "epoch": 60.0,
      "grad_norm": 0.5702971816062927,
      "learning_rate": 2e-05,
      "loss": 0.087,
      "step": 6360
    },
    {
      "epoch": 60.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7738095238095238,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6989247311827957,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.9006622516556291,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8553459119496856,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8148148148148148,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7857142857142857,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13854016363620758,
      "eval_overall_accuracy": 0.9594696022016512,
      "eval_overall_f1": 0.8240109140518419,
      "eval_overall_precision": 0.766497461928934,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.3704,
      "eval_samples_per_second": 504.86,
      "eval_steps_per_second": 8.099,
      "step": 6360
    },
    {
      "epoch": 61.0,
      "grad_norm": 1.0941648483276367,
      "learning_rate": 1.9500000000000003e-05,
      "loss": 0.0836,
      "step": 6466
    },
    {
      "epoch": 61.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7745664739884393,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6836734693877551,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.9108910891089109,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8625,
      "eval_PERSON_recall": 0.965034965034965,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14069223403930664,
      "eval_overall_accuracy": 0.9602201651238429,
      "eval_overall_f1": 0.827027027027027,
      "eval_overall_precision": 0.7630922693266833,
      "eval_overall_recall": 0.9026548672566371,
      "eval_runtime": 0.3687,
      "eval_samples_per_second": 507.138,
      "eval_steps_per_second": 8.136,
      "step": 6466
    },
    {
      "epoch": 62.0,
      "grad_norm": 0.9212659597396851,
      "learning_rate": 1.9e-05,
      "loss": 0.0823,
      "step": 6572
    },
    {
      "epoch": 62.0,
      "eval_LOCATION_f1": 0.8026315789473685,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7625,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7710843373493976,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7032967032967034,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6415094339622641,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5666666666666667,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13715825974941254,
      "eval_overall_accuracy": 0.9602201651238429,
      "eval_overall_f1": 0.823045267489712,
      "eval_overall_precision": 0.7692307692307693,
      "eval_overall_recall": 0.8849557522123894,
      "eval_runtime": 0.3668,
      "eval_samples_per_second": 509.835,
      "eval_steps_per_second": 8.179,
      "step": 6572
    },
    {
      "epoch": 63.0,
      "grad_norm": 0.8010810613632202,
      "learning_rate": 1.85e-05,
      "loss": 0.0814,
      "step": 6678
    },
    {
      "epoch": 63.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7674418604651163,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6804123711340206,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7272727272727274,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6896551724137931,
      "eval_TIME_recall": 0.7692307692307693,
      "eval_loss": 0.1415587216615677,
      "eval_overall_accuracy": 0.957968476357268,
      "eval_overall_f1": 0.814614343707713,
      "eval_overall_precision": 0.7525,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3713,
      "eval_samples_per_second": 503.694,
      "eval_steps_per_second": 8.081,
      "step": 6678
    },
    {
      "epoch": 64.0,
      "grad_norm": 1.7902244329452515,
      "learning_rate": 1.8e-05,
      "loss": 0.0806,
      "step": 6784
    },
    {
      "epoch": 64.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7486033519553071,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6442307692307693,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.9078947368421052,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8571428571428571,
      "eval_PERSON_recall": 0.965034965034965,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7142857142857142,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6666666666666666,
      "eval_TIME_recall": 0.7692307692307693,
      "eval_loss": 0.1516939103603363,
      "eval_overall_accuracy": 0.9564673505128847,
      "eval_overall_f1": 0.8155080213903745,
      "eval_overall_precision": 0.7457212713936431,
      "eval_overall_recall": 0.8997050147492626,
      "eval_runtime": 0.372,
      "eval_samples_per_second": 502.674,
      "eval_steps_per_second": 8.064,
      "step": 6784
    },
    {
      "epoch": 65.0,
      "grad_norm": 0.6327823400497437,
      "learning_rate": 1.75e-05,
      "loss": 0.0794,
      "step": 6890
    },
    {
      "epoch": 65.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7784431137724551,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7065217391304348,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14160560071468353,
      "eval_overall_accuracy": 0.9589692269201902,
      "eval_overall_f1": 0.8201634877384195,
      "eval_overall_precision": 0.7620253164556962,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3687,
      "eval_samples_per_second": 507.195,
      "eval_steps_per_second": 8.137,
      "step": 6890
    },
    {
      "epoch": 66.0,
      "grad_norm": 0.9762814044952393,
      "learning_rate": 1.7000000000000003e-05,
      "loss": 0.079,
      "step": 6996
    },
    {
      "epoch": 66.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7861271676300577,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6938775510204082,
      "eval_ORGANIZATION_recall": 0.9066666666666666,
      "eval_PERSON_f1": 0.9078947368421052,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8571428571428571,
      "eval_PERSON_recall": 0.965034965034965,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1456669718027115,
      "eval_overall_accuracy": 0.9592194145609206,
      "eval_overall_f1": 0.8286099865047233,
      "eval_overall_precision": 0.763681592039801,
      "eval_overall_recall": 0.9056047197640118,
      "eval_runtime": 0.3699,
      "eval_samples_per_second": 505.498,
      "eval_steps_per_second": 8.11,
      "step": 6996
    },
    {
      "epoch": 67.0,
      "grad_norm": 0.4873846173286438,
      "learning_rate": 1.65e-05,
      "loss": 0.0811,
      "step": 7102
    },
    {
      "epoch": 67.0,
      "eval_LOCATION_f1": 0.8104575163398693,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7654320987654321,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7738095238095238,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6989247311827957,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.9006622516556291,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8553459119496856,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.13851337134838104,
      "eval_overall_accuracy": 0.9604703527645734,
      "eval_overall_f1": 0.8267394270122784,
      "eval_overall_precision": 0.7690355329949239,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3724,
      "eval_samples_per_second": 502.102,
      "eval_steps_per_second": 8.055,
      "step": 7102
    },
    {
      "epoch": 68.0,
      "grad_norm": 0.9525014758110046,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.0792,
      "step": 7208
    },
    {
      "epoch": 68.0,
      "eval_LOCATION_f1": 0.7870967741935484,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7349397590361446,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7570621468926553,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6568627450980392,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.9072847682119206,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8616352201257862,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14804454147815704,
      "eval_overall_accuracy": 0.9572179134350763,
      "eval_overall_f1": 0.814516129032258,
      "eval_overall_precision": 0.7481481481481481,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3712,
      "eval_samples_per_second": 503.718,
      "eval_steps_per_second": 8.081,
      "step": 7208
    },
    {
      "epoch": 69.0,
      "grad_norm": 2.1547799110412598,
      "learning_rate": 1.55e-05,
      "loss": 0.0795,
      "step": 7314
    },
    {
      "epoch": 69.0,
      "eval_LOCATION_f1": 0.8,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7469879518072289,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7701149425287357,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6767676767676768,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.8918032786885245,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8395061728395061,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.1455415040254593,
      "eval_overall_accuracy": 0.9584688516387291,
      "eval_overall_f1": 0.8198924731182796,
      "eval_overall_precision": 0.7530864197530864,
      "eval_overall_recall": 0.8997050147492626,
      "eval_runtime": 0.3711,
      "eval_samples_per_second": 503.946,
      "eval_steps_per_second": 8.085,
      "step": 7314
    },
    {
      "epoch": 70.0,
      "grad_norm": 1.8800172805786133,
      "learning_rate": 1.5e-05,
      "loss": 0.0805,
      "step": 7420
    },
    {
      "epoch": 70.0,
      "eval_LOCATION_f1": 0.8104575163398693,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7654320987654321,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7738095238095238,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6989247311827957,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.9013157894736843,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8509316770186336,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.14219874143600464,
      "eval_overall_accuracy": 0.9599699774831123,
      "eval_overall_f1": 0.8272108843537416,
      "eval_overall_precision": 0.7676767676767676,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.3708,
      "eval_samples_per_second": 504.365,
      "eval_steps_per_second": 8.091,
      "step": 7420
    },
    {
      "epoch": 71.0,
      "grad_norm": 2.311293363571167,
      "learning_rate": 1.45e-05,
      "loss": 0.0809,
      "step": 7526
    },
    {
      "epoch": 71.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.788235294117647,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7052631578947368,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.9013157894736843,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8509316770186336,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14474977552890778,
      "eval_overall_accuracy": 0.9594696022016512,
      "eval_overall_f1": 0.8265582655826559,
      "eval_overall_precision": 0.7644110275689223,
      "eval_overall_recall": 0.8997050147492626,
      "eval_runtime": 0.3706,
      "eval_samples_per_second": 504.553,
      "eval_steps_per_second": 8.094,
      "step": 7526
    },
    {
      "epoch": 72.0,
      "grad_norm": 1.1506495475769043,
      "learning_rate": 1.4000000000000001e-05,
      "loss": 0.0787,
      "step": 7632
    },
    {
      "epoch": 72.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7976190476190477,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7204301075268817,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.9013157894736843,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8509316770186336,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1447576880455017,
      "eval_overall_accuracy": 0.9597197898423818,
      "eval_overall_f1": 0.8233695652173914,
      "eval_overall_precision": 0.7632241813602015,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3738,
      "eval_samples_per_second": 500.28,
      "eval_steps_per_second": 8.026,
      "step": 7632
    },
    {
      "epoch": 73.0,
      "grad_norm": 1.78630793094635,
      "learning_rate": 1.3500000000000001e-05,
      "loss": 0.0792,
      "step": 7738
    },
    {
      "epoch": 73.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7590361445783131,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6923076923076923,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.9013157894736843,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8509316770186336,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6792452830188679,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.6,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.8,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7586206896551724,
      "eval_TIME_recall": 0.8461538461538461,
      "eval_loss": 0.14058354496955872,
      "eval_overall_accuracy": 0.9602201651238429,
      "eval_overall_f1": 0.8224043715846995,
      "eval_overall_precision": 0.7659033078880407,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3699,
      "eval_samples_per_second": 505.587,
      "eval_steps_per_second": 8.111,
      "step": 7738
    },
    {
      "epoch": 74.0,
      "grad_norm": 1.3375426530838013,
      "learning_rate": 1.3000000000000001e-05,
      "loss": 0.078,
      "step": 7844
    },
    {
      "epoch": 74.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7857142857142856,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7096774193548387,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14712856709957123,
      "eval_overall_accuracy": 0.9582186639979985,
      "eval_overall_f1": 0.8179347826086957,
      "eval_overall_precision": 0.7581863979848866,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3703,
      "eval_samples_per_second": 505.025,
      "eval_steps_per_second": 8.102,
      "step": 7844
    },
    {
      "epoch": 75.0,
      "grad_norm": 1.585189938545227,
      "learning_rate": 1.25e-05,
      "loss": 0.0773,
      "step": 7950
    },
    {
      "epoch": 75.0,
      "eval_LOCATION_f1": 0.7922077922077924,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7439024390243902,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.783625730994152,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6979166666666666,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.9013157894736843,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8509316770186336,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1459941565990448,
      "eval_overall_accuracy": 0.9589692269201902,
      "eval_overall_f1": 0.8227334235453315,
      "eval_overall_precision": 0.76,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.3706,
      "eval_samples_per_second": 504.644,
      "eval_steps_per_second": 8.096,
      "step": 7950
    },
    {
      "epoch": 76.0,
      "grad_norm": 1.1429756879806519,
      "learning_rate": 1.2e-05,
      "loss": 0.0772,
      "step": 8056
    },
    {
      "epoch": 76.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.783625730994152,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6979166666666666,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.9042904290429041,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85625,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14840912818908691,
      "eval_overall_accuracy": 0.9584688516387291,
      "eval_overall_f1": 0.8265582655826559,
      "eval_overall_precision": 0.7644110275689223,
      "eval_overall_recall": 0.8997050147492626,
      "eval_runtime": 0.3692,
      "eval_samples_per_second": 506.499,
      "eval_steps_per_second": 8.126,
      "step": 8056
    },
    {
      "epoch": 77.0,
      "grad_norm": 1.014310359954834,
      "learning_rate": 1.1500000000000002e-05,
      "loss": 0.0767,
      "step": 8162
    },
    {
      "epoch": 77.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7790697674418604,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6907216494845361,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.9072847682119206,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8616352201257862,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.15152573585510254,
      "eval_overall_accuracy": 0.9574681010758068,
      "eval_overall_f1": 0.8227334235453315,
      "eval_overall_precision": 0.76,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.3708,
      "eval_samples_per_second": 504.318,
      "eval_steps_per_second": 8.091,
      "step": 8162
    },
    {
      "epoch": 78.0,
      "grad_norm": 1.17866051197052,
      "learning_rate": 1.1000000000000001e-05,
      "loss": 0.0757,
      "step": 8268
    },
    {
      "epoch": 78.0,
      "eval_LOCATION_f1": 0.8289473684210527,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7875,
      "eval_LOCATION_recall": 0.875,
      "eval_ORGANIZATION_f1": 0.7682926829268292,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7078651685393258,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8970099667774087,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8544303797468354,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14352728426456451,
      "eval_overall_accuracy": 0.9589692269201902,
      "eval_overall_f1": 0.8253094910591473,
      "eval_overall_precision": 0.7731958762886598,
      "eval_overall_recall": 0.8849557522123894,
      "eval_runtime": 0.374,
      "eval_samples_per_second": 500.053,
      "eval_steps_per_second": 8.022,
      "step": 8268
    },
    {
      "epoch": 79.0,
      "grad_norm": 2.2889716625213623,
      "learning_rate": 1.05e-05,
      "loss": 0.0767,
      "step": 8374
    },
    {
      "epoch": 79.0,
      "eval_LOCATION_f1": 0.8104575163398693,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7654320987654321,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7904191616766466,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.717391304347826,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.9042904290429041,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85625,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14561718702316284,
      "eval_overall_accuracy": 0.9602201651238429,
      "eval_overall_f1": 0.8267394270122784,
      "eval_overall_precision": 0.7690355329949239,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3708,
      "eval_samples_per_second": 504.277,
      "eval_steps_per_second": 8.09,
      "step": 8374
    },
    {
      "epoch": 80.0,
      "grad_norm": 0.5872527956962585,
      "learning_rate": 1e-05,
      "loss": 0.0752,
      "step": 8480
    },
    {
      "epoch": 80.0,
      "eval_LOCATION_f1": 0.823529411764706,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7777777777777778,
      "eval_LOCATION_recall": 0.875,
      "eval_ORGANIZATION_f1": 0.7570621468926553,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6568627450980392,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.9042904290429041,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85625,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7142857142857142,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6666666666666666,
      "eval_TIME_recall": 0.7692307692307693,
      "eval_loss": 0.15562912821769714,
      "eval_overall_accuracy": 0.9554665999499625,
      "eval_overall_f1": 0.8161073825503355,
      "eval_overall_precision": 0.7487684729064039,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.3781,
      "eval_samples_per_second": 494.629,
      "eval_steps_per_second": 7.935,
      "step": 8480
    },
    {
      "epoch": 81.0,
      "grad_norm": 1.4843671321868896,
      "learning_rate": 9.5e-06,
      "loss": 0.078,
      "step": 8586
    },
    {
      "epoch": 81.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7810650887573964,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7021276595744681,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.9013157894736843,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8509316770186336,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.75,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14394411444664001,
      "eval_overall_accuracy": 0.9599699774831123,
      "eval_overall_f1": 0.8238482384823848,
      "eval_overall_precision": 0.7619047619047619,
      "eval_overall_recall": 0.8967551622418879,
      "eval_runtime": 0.3729,
      "eval_samples_per_second": 501.415,
      "eval_steps_per_second": 8.044,
      "step": 8586
    },
    {
      "epoch": 82.0,
      "grad_norm": 1.6159791946411133,
      "learning_rate": 9e-06,
      "loss": 0.0733,
      "step": 8692
    },
    {
      "epoch": 82.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.788235294117647,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7052631578947368,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7142857142857142,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.6666666666666666,
      "eval_TIME_recall": 0.7692307692307693,
      "eval_loss": 0.14782680571079254,
      "eval_overall_accuracy": 0.9587190392794596,
      "eval_overall_f1": 0.817320703653586,
      "eval_overall_precision": 0.755,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.3733,
      "eval_samples_per_second": 500.893,
      "eval_steps_per_second": 8.036,
      "step": 8692
    },
    {
      "epoch": 83.0,
      "grad_norm": 1.871289610862732,
      "learning_rate": 8.500000000000002e-06,
      "loss": 0.0747,
      "step": 8798
    },
    {
      "epoch": 83.0,
      "eval_LOCATION_f1": 0.7973856209150327,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7530864197530864,
      "eval_LOCATION_recall": 0.8472222222222222,
      "eval_ORGANIZATION_f1": 0.7831325301204819,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7142857142857143,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14479823410511017,
      "eval_overall_accuracy": 0.9599699774831123,
      "eval_overall_f1": 0.8224043715846995,
      "eval_overall_precision": 0.7659033078880407,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3747,
      "eval_samples_per_second": 499.023,
      "eval_steps_per_second": 8.006,
      "step": 8798
    },
    {
      "epoch": 84.0,
      "grad_norm": 2.436727285385132,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.0755,
      "step": 8904
    },
    {
      "epoch": 84.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7928994082840236,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7127659574468085,
      "eval_ORGANIZATION_recall": 0.8933333333333333,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7777777777777779,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.75,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14651836454868317,
      "eval_overall_accuracy": 0.9597197898423818,
      "eval_overall_f1": 0.8233695652173914,
      "eval_overall_precision": 0.7632241813602015,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3731,
      "eval_samples_per_second": 501.201,
      "eval_steps_per_second": 8.041,
      "step": 8904
    },
    {
      "epoch": 85.0,
      "grad_norm": 1.8641331195831299,
      "learning_rate": 7.5e-06,
      "loss": 0.0757,
      "step": 9010
    },
    {
      "epoch": 85.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7682926829268292,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7078651685393258,
      "eval_ORGANIZATION_recall": 0.84,
      "eval_PERSON_f1": 0.8881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8385093167701864,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14340606331825256,
      "eval_overall_accuracy": 0.9592194145609206,
      "eval_overall_f1": 0.8169398907103824,
      "eval_overall_precision": 0.7608142493638677,
      "eval_overall_recall": 0.8820058997050148,
      "eval_runtime": 0.3727,
      "eval_samples_per_second": 501.795,
      "eval_steps_per_second": 8.05,
      "step": 9010
    },
    {
      "epoch": 86.0,
      "grad_norm": 2.203537702560425,
      "learning_rate": 7.000000000000001e-06,
      "loss": 0.076,
      "step": 9116
    },
    {
      "epoch": 86.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7710843373493976,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7032967032967034,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6181818181818182,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.53125,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1468215435743332,
      "eval_overall_accuracy": 0.9587190392794596,
      "eval_overall_f1": 0.8185538881309687,
      "eval_overall_precision": 0.7614213197969543,
      "eval_overall_recall": 0.8849557522123894,
      "eval_runtime": 0.3717,
      "eval_samples_per_second": 503.095,
      "eval_steps_per_second": 8.071,
      "step": 9116
    },
    {
      "epoch": 87.0,
      "grad_norm": 1.147968053817749,
      "learning_rate": 6.5000000000000004e-06,
      "loss": 0.0721,
      "step": 9222
    },
    {
      "epoch": 87.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7710843373493976,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7032967032967034,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8947368421052632,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84472049689441,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7777777777777779,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.75,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14543673396110535,
      "eval_overall_accuracy": 0.9592194145609206,
      "eval_overall_f1": 0.8212824010914053,
      "eval_overall_precision": 0.7639593908629442,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3713,
      "eval_samples_per_second": 503.66,
      "eval_steps_per_second": 8.08,
      "step": 9222
    },
    {
      "epoch": 88.0,
      "grad_norm": 1.8473503589630127,
      "learning_rate": 6e-06,
      "loss": 0.0724,
      "step": 9328
    },
    {
      "epoch": 88.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7784431137724551,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7065217391304348,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6666666666666667,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5806451612903226,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14697220921516418,
      "eval_overall_accuracy": 0.9594696022016512,
      "eval_overall_f1": 0.8240109140518419,
      "eval_overall_precision": 0.766497461928934,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.3718,
      "eval_samples_per_second": 502.894,
      "eval_steps_per_second": 8.068,
      "step": 9328
    },
    {
      "epoch": 89.0,
      "grad_norm": 1.6724920272827148,
      "learning_rate": 5.500000000000001e-06,
      "loss": 0.074,
      "step": 9434
    },
    {
      "epoch": 89.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7738095238095238,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6989247311827957,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6545454545454547,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5625,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14885887503623962,
      "eval_overall_accuracy": 0.9589692269201902,
      "eval_overall_f1": 0.8217687074829931,
      "eval_overall_precision": 0.7626262626262627,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.373,
      "eval_samples_per_second": 501.288,
      "eval_steps_per_second": 8.042,
      "step": 9434
    },
    {
      "epoch": 90.0,
      "grad_norm": 1.6046056747436523,
      "learning_rate": 5e-06,
      "loss": 0.0721,
      "step": 9540
    },
    {
      "epoch": 90.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7810650887573964,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7021276595744681,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8976897689768978,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.85,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14962825179100037,
      "eval_overall_accuracy": 0.9587190392794596,
      "eval_overall_f1": 0.819538670284939,
      "eval_overall_precision": 0.7587939698492462,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.3743,
      "eval_samples_per_second": 499.541,
      "eval_steps_per_second": 8.014,
      "step": 9540
    },
    {
      "epoch": 91.0,
      "grad_norm": 0.5925461649894714,
      "learning_rate": 4.5e-06,
      "loss": 0.0766,
      "step": 9646
    },
    {
      "epoch": 91.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7664670658682634,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6956521739130435,
      "eval_ORGANIZATION_recall": 0.8533333333333334,
      "eval_PERSON_f1": 0.8881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8385093167701864,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14539198577404022,
      "eval_overall_accuracy": 0.9587190392794596,
      "eval_overall_f1": 0.8125000000000001,
      "eval_overall_precision": 0.7531486146095718,
      "eval_overall_recall": 0.8820058997050148,
      "eval_runtime": 0.3743,
      "eval_samples_per_second": 499.533,
      "eval_steps_per_second": 8.014,
      "step": 9646
    },
    {
      "epoch": 92.0,
      "grad_norm": 2.1363561153411865,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.075,
      "step": 9752
    },
    {
      "epoch": 92.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7810650887573964,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7021276595744681,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8947368421052632,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84472049689441,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1483411341905594,
      "eval_overall_accuracy": 0.9587190392794596,
      "eval_overall_f1": 0.8184281842818428,
      "eval_overall_precision": 0.7568922305764411,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.3707,
      "eval_samples_per_second": 504.407,
      "eval_steps_per_second": 8.092,
      "step": 9752
    },
    {
      "epoch": 93.0,
      "grad_norm": 0.5211588144302368,
      "learning_rate": 3.5000000000000004e-06,
      "loss": 0.0713,
      "step": 9858
    },
    {
      "epoch": 93.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7810650887573964,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7021276595744681,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8910891089108911,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84375,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14794917404651642,
      "eval_overall_accuracy": 0.9587190392794596,
      "eval_overall_f1": 0.8168249660786975,
      "eval_overall_precision": 0.7562814070351759,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.374,
      "eval_samples_per_second": 500.047,
      "eval_steps_per_second": 8.022,
      "step": 9858
    },
    {
      "epoch": 94.0,
      "grad_norm": 0.8026951551437378,
      "learning_rate": 3e-06,
      "loss": 0.0727,
      "step": 9964
    },
    {
      "epoch": 94.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7810650887573964,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7021276595744681,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.9013157894736843,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8509316770186336,
      "eval_PERSON_recall": 0.958041958041958,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.1480274349451065,
      "eval_overall_accuracy": 0.9589692269201902,
      "eval_overall_f1": 0.8211382113821138,
      "eval_overall_precision": 0.7593984962406015,
      "eval_overall_recall": 0.8938053097345132,
      "eval_runtime": 0.3712,
      "eval_samples_per_second": 503.809,
      "eval_steps_per_second": 8.083,
      "step": 9964
    },
    {
      "epoch": 95.0,
      "grad_norm": 0.5980091691017151,
      "learning_rate": 2.5e-06,
      "loss": 0.0726,
      "step": 10070
    },
    {
      "epoch": 95.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7857142857142856,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7096774193548387,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8947368421052632,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.84472049689441,
      "eval_PERSON_recall": 0.951048951048951,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14869406819343567,
      "eval_overall_accuracy": 0.9587190392794596,
      "eval_overall_f1": 0.819538670284939,
      "eval_overall_precision": 0.7587939698492462,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.3746,
      "eval_samples_per_second": 499.2,
      "eval_steps_per_second": 8.009,
      "step": 10070
    },
    {
      "epoch": 96.0,
      "grad_norm": 0.48421594500541687,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0733,
      "step": 10176
    },
    {
      "epoch": 96.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7810650887573964,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7021276595744681,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8844884488448846,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8375,
      "eval_PERSON_recall": 0.9370629370629371,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14902207255363464,
      "eval_overall_accuracy": 0.957968476357268,
      "eval_overall_f1": 0.814111261872456,
      "eval_overall_precision": 0.7537688442211056,
      "eval_overall_recall": 0.8849557522123894,
      "eval_runtime": 0.3742,
      "eval_samples_per_second": 499.722,
      "eval_steps_per_second": 8.017,
      "step": 10176
    },
    {
      "epoch": 97.0,
      "grad_norm": 0.8765180110931396,
      "learning_rate": 1.5e-06,
      "loss": 0.0721,
      "step": 10282
    },
    {
      "epoch": 97.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7692307692307693,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.6914893617021277,
      "eval_ORGANIZATION_recall": 0.8666666666666667,
      "eval_PERSON_f1": 0.8881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8385093167701864,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14730840921401978,
      "eval_overall_accuracy": 0.9584688516387291,
      "eval_overall_f1": 0.8157181571815718,
      "eval_overall_precision": 0.7543859649122807,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3716,
      "eval_samples_per_second": 503.173,
      "eval_steps_per_second": 8.072,
      "step": 10282
    },
    {
      "epoch": 98.0,
      "grad_norm": 0.884979784488678,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0722,
      "step": 10388
    },
    {
      "epoch": 98.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7810650887573964,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7021276595744681,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8385093167701864,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14785435795783997,
      "eval_overall_accuracy": 0.9589692269201902,
      "eval_overall_f1": 0.8184281842818428,
      "eval_overall_precision": 0.7568922305764411,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.3715,
      "eval_samples_per_second": 503.38,
      "eval_steps_per_second": 8.076,
      "step": 10388
    },
    {
      "epoch": 99.0,
      "grad_norm": 1.1624181270599365,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0729,
      "step": 10494
    },
    {
      "epoch": 99.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7810650887573964,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7021276595744681,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8385093167701864,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6071428571428571,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5151515151515151,
      "eval_QUANTITY_recall": 0.7391304347826086,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14793308079242706,
      "eval_overall_accuracy": 0.9587190392794596,
      "eval_overall_f1": 0.8157181571815718,
      "eval_overall_precision": 0.7543859649122807,
      "eval_overall_recall": 0.887905604719764,
      "eval_runtime": 0.3726,
      "eval_samples_per_second": 501.877,
      "eval_steps_per_second": 8.051,
      "step": 10494
    },
    {
      "epoch": 100.0,
      "grad_norm": 1.330012559890747,
      "learning_rate": 0.0,
      "loss": 0.0718,
      "step": 10600
    },
    {
      "epoch": 100.0,
      "eval_LOCATION_f1": 0.8051948051948052,
      "eval_LOCATION_number": 72,
      "eval_LOCATION_precision": 0.7560975609756098,
      "eval_LOCATION_recall": 0.8611111111111112,
      "eval_ORGANIZATION_f1": 0.7810650887573964,
      "eval_ORGANIZATION_number": 75,
      "eval_ORGANIZATION_precision": 0.7021276595744681,
      "eval_ORGANIZATION_recall": 0.88,
      "eval_PERSON_f1": 0.8881578947368421,
      "eval_PERSON_number": 143,
      "eval_PERSON_precision": 0.8385093167701864,
      "eval_PERSON_recall": 0.9440559440559441,
      "eval_QUANTITY_f1": 0.6428571428571429,
      "eval_QUANTITY_number": 23,
      "eval_QUANTITY_precision": 0.5454545454545454,
      "eval_QUANTITY_recall": 0.782608695652174,
      "eval_TIME_f1": 0.7636363636363636,
      "eval_TIME_number": 26,
      "eval_TIME_precision": 0.7241379310344828,
      "eval_TIME_recall": 0.8076923076923077,
      "eval_loss": 0.14785712957382202,
      "eval_overall_accuracy": 0.9589692269201902,
      "eval_overall_f1": 0.8184281842818428,
      "eval_overall_precision": 0.7568922305764411,
      "eval_overall_recall": 0.8908554572271387,
      "eval_runtime": 0.3703,
      "eval_samples_per_second": 504.948,
      "eval_steps_per_second": 8.101,
      "step": 10600
    },
    {
      "epoch": 100.0,
      "step": 10600,
      "total_flos": 4533533359897542.0,
      "train_loss": 0.12740072155898471,
      "train_runtime": 632.9493,
      "train_samples_per_second": 266.53,
      "train_steps_per_second": 16.747
    }
  ],
  "logging_steps": 500,
  "max_steps": 10600,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 100,
  "save_steps": 500,
  "total_flos": 4533533359897542.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}