{ "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": 2.2747724056243896, "learning_rate": 4.9500000000000004e-05, "loss": 1.1647, "step": 106 }, { "epoch": 1.0, "eval_LOCATION_f1": 0.0, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.0, "eval_LOCATION_recall": 0.0, "eval_ORGANIZATION_f1": 0.0, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.0, "eval_ORGANIZATION_recall": 0.0, "eval_PERSON_f1": 0.0, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.0, "eval_PERSON_recall": 0.0, "eval_QUANTITY_f1": 0.0, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.0, "eval_QUANTITY_recall": 0.0, "eval_TIME_f1": 0.0, "eval_TIME_number": 28, "eval_TIME_precision": 0.0, "eval_TIME_recall": 0.0, "eval_loss": 0.7061953544616699, "eval_overall_accuracy": 0.8409311348205626, "eval_overall_f1": 0.0, "eval_overall_precision": 0.0, "eval_overall_recall": 0.0, "eval_runtime": 0.3721, "eval_samples_per_second": 502.553, "eval_steps_per_second": 8.062, "step": 106 }, { "epoch": 2.0, "grad_norm": 0.9771762490272522, "learning_rate": 4.9e-05, "loss": 0.6874, "step": 212 }, { "epoch": 2.0, "eval_LOCATION_f1": 0.0, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.0, "eval_LOCATION_recall": 0.0, "eval_ORGANIZATION_f1": 0.0, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.0, "eval_ORGANIZATION_recall": 0.0, "eval_PERSON_f1": 0.012903225806451615, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.2, "eval_PERSON_recall": 0.006666666666666667, "eval_QUANTITY_f1": 0.0, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.0, "eval_QUANTITY_recall": 0.0, "eval_TIME_f1": 0.0, "eval_TIME_number": 28, "eval_TIME_precision": 0.0, "eval_TIME_recall": 0.0, "eval_loss": 0.5869600176811218, "eval_overall_accuracy": 0.8426285160038798, "eval_overall_f1": 0.005633802816901408, "eval_overall_precision": 0.125, "eval_overall_recall": 0.002881844380403458, "eval_runtime": 0.3708, "eval_samples_per_second": 504.33, "eval_steps_per_second": 8.091, "step": 212 }, { "epoch": 3.0, "grad_norm": 0.8677288293838501, "learning_rate": 4.85e-05, "loss": 0.5688, "step": 318 }, { "epoch": 3.0, "eval_LOCATION_f1": 0.07228915662650602, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.25, "eval_LOCATION_recall": 0.04225352112676056, "eval_ORGANIZATION_f1": 0.025974025974025976, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.08333333333333333, "eval_ORGANIZATION_recall": 0.015384615384615385, "eval_PERSON_f1": 0.19213973799126635, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.27848101265822783, "eval_PERSON_recall": 0.14666666666666667, "eval_QUANTITY_f1": 0.0, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.0, "eval_QUANTITY_recall": 0.0, "eval_TIME_f1": 0.04878048780487805, "eval_TIME_number": 28, "eval_TIME_precision": 0.07692307692307693, "eval_TIME_recall": 0.03571428571428571, "eval_loss": 0.4580833613872528, "eval_overall_accuracy": 0.8642095053346266, "eval_overall_f1": 0.11663066954643629, "eval_overall_precision": 0.23275862068965517, "eval_overall_recall": 0.07780979827089338, "eval_runtime": 0.3745, "eval_samples_per_second": 499.319, "eval_steps_per_second": 8.01, "step": 318 }, { "epoch": 4.0, "grad_norm": 1.8199503421783447, "learning_rate": 4.8e-05, "loss": 0.4433, "step": 424 }, { "epoch": 4.0, "eval_LOCATION_f1": 0.3333333333333333, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.40816326530612246, "eval_LOCATION_recall": 0.28169014084507044, "eval_ORGANIZATION_f1": 0.14893617021276598, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.2413793103448276, "eval_ORGANIZATION_recall": 0.1076923076923077, "eval_PERSON_f1": 0.54320987654321, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.5057471264367817, "eval_PERSON_recall": 0.5866666666666667, "eval_QUANTITY_f1": 0.0784313725490196, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.1111111111111111, "eval_QUANTITY_recall": 0.06060606060606061, "eval_TIME_f1": 0.3137254901960784, "eval_TIME_number": 28, "eval_TIME_precision": 0.34782608695652173, "eval_TIME_recall": 0.2857142857142857, "eval_loss": 0.3302854299545288, "eval_overall_accuracy": 0.9020368574199806, "eval_overall_f1": 0.390625, "eval_overall_precision": 0.42662116040955633, "eval_overall_recall": 0.36023054755043227, "eval_runtime": 0.3786, "eval_samples_per_second": 493.975, "eval_steps_per_second": 7.925, "step": 424 }, { "epoch": 5.0, "grad_norm": 1.0272399187088013, "learning_rate": 4.75e-05, "loss": 0.3385, "step": 530 }, { "epoch": 5.0, "eval_LOCATION_f1": 0.5548387096774193, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.5119047619047619, "eval_LOCATION_recall": 0.6056338028169014, "eval_ORGANIZATION_f1": 0.4305555555555556, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.3924050632911392, "eval_ORGANIZATION_recall": 0.47692307692307695, "eval_PERSON_f1": 0.7176470588235295, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.6421052631578947, "eval_PERSON_recall": 0.8133333333333334, "eval_QUANTITY_f1": 0.411764705882353, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.4, "eval_QUANTITY_recall": 0.42424242424242425, "eval_TIME_f1": 0.6071428571428571, "eval_TIME_number": 28, "eval_TIME_precision": 0.6071428571428571, "eval_TIME_recall": 0.6071428571428571, "eval_loss": 0.2530401051044464, "eval_overall_accuracy": 0.9274975751697381, "eval_overall_f1": 0.5950196592398428, "eval_overall_precision": 0.5456730769230769, "eval_overall_recall": 0.654178674351585, "eval_runtime": 0.3801, "eval_samples_per_second": 492.026, "eval_steps_per_second": 7.893, "step": 530 }, { "epoch": 6.0, "grad_norm": 1.1298264265060425, "learning_rate": 4.7e-05, "loss": 0.2744, "step": 636 }, { "epoch": 6.0, "eval_LOCATION_f1": 0.6506024096385542, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.5684210526315789, "eval_LOCATION_recall": 0.7605633802816901, "eval_ORGANIZATION_f1": 0.5212121212121211, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.43, "eval_ORGANIZATION_recall": 0.6615384615384615, "eval_PERSON_f1": 0.7768115942028986, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.6871794871794872, "eval_PERSON_recall": 0.8933333333333333, "eval_QUANTITY_f1": 0.47368421052631576, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.4186046511627907, "eval_QUANTITY_recall": 0.5454545454545454, "eval_TIME_f1": 0.6451612903225806, "eval_TIME_number": 28, "eval_TIME_precision": 0.5882352941176471, "eval_TIME_recall": 0.7142857142857143, "eval_loss": 0.23107779026031494, "eval_overall_accuracy": 0.9342870999030067, "eval_overall_f1": 0.660933660933661, "eval_overall_precision": 0.576017130620985, "eval_overall_recall": 0.7752161383285303, "eval_runtime": 0.3766, "eval_samples_per_second": 496.553, "eval_steps_per_second": 7.966, "step": 636 }, { "epoch": 7.0, "grad_norm": 2.4953019618988037, "learning_rate": 4.6500000000000005e-05, "loss": 0.2399, "step": 742 }, { "epoch": 7.0, "eval_LOCATION_f1": 0.6746987951807228, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.5894736842105263, "eval_LOCATION_recall": 0.7887323943661971, "eval_ORGANIZATION_f1": 0.5714285714285713, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.4943820224719101, "eval_ORGANIZATION_recall": 0.676923076923077, "eval_PERSON_f1": 0.8143712574850299, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.7391304347826086, "eval_PERSON_recall": 0.9066666666666666, "eval_QUANTITY_f1": 0.5263157894736842, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.46511627906976744, "eval_QUANTITY_recall": 0.6060606060606061, "eval_TIME_f1": 0.6451612903225806, "eval_TIME_number": 28, "eval_TIME_precision": 0.5882352941176471, "eval_TIME_recall": 0.7142857142857143, "eval_loss": 0.19521614909172058, "eval_overall_accuracy": 0.9401066925315228, "eval_overall_f1": 0.696969696969697, "eval_overall_precision": 0.6202247191011236, "eval_overall_recall": 0.7953890489913544, "eval_runtime": 0.378, "eval_samples_per_second": 494.72, "eval_steps_per_second": 7.937, "step": 742 }, { "epoch": 8.0, "grad_norm": 0.9211530685424805, "learning_rate": 4.600000000000001e-05, "loss": 0.2121, "step": 848 }, { "epoch": 8.0, "eval_LOCATION_f1": 0.6705882352941176, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.5757575757575758, "eval_LOCATION_recall": 0.8028169014084507, "eval_ORGANIZATION_f1": 0.6091954022988506, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.48623853211009177, "eval_ORGANIZATION_recall": 0.8153846153846154, "eval_PERSON_f1": 0.8217522658610271, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.7513812154696132, "eval_PERSON_recall": 0.9066666666666666, "eval_QUANTITY_f1": 0.5641025641025641, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.4888888888888889, "eval_QUANTITY_recall": 0.6666666666666666, "eval_TIME_f1": 0.6984126984126985, "eval_TIME_number": 28, "eval_TIME_precision": 0.6285714285714286, "eval_TIME_recall": 0.7857142857142857, "eval_loss": 0.1916876584291458, "eval_overall_accuracy": 0.9381668283220175, "eval_overall_f1": 0.7107843137254902, "eval_overall_precision": 0.6183368869936035, "eval_overall_recall": 0.8357348703170029, "eval_runtime": 0.3809, "eval_samples_per_second": 490.906, "eval_steps_per_second": 7.875, "step": 848 }, { "epoch": 9.0, "grad_norm": 2.266162395477295, "learning_rate": 4.55e-05, "loss": 0.196, "step": 954 }, { "epoch": 9.0, "eval_LOCATION_f1": 0.7124999999999999, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.6404494382022472, "eval_LOCATION_recall": 0.8028169014084507, "eval_ORGANIZATION_f1": 0.6490066225165563, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5697674418604651, "eval_ORGANIZATION_recall": 0.7538461538461538, "eval_PERSON_f1": 0.8447204968944099, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.7906976744186046, "eval_PERSON_recall": 0.9066666666666666, "eval_QUANTITY_f1": 0.5753424657534246, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.525, "eval_QUANTITY_recall": 0.6363636363636364, "eval_TIME_f1": 0.7118644067796611, "eval_TIME_number": 28, "eval_TIME_precision": 0.6774193548387096, "eval_TIME_recall": 0.75, "eval_loss": 0.1588808298110962, "eval_overall_accuracy": 0.9473811833171678, "eval_overall_f1": 0.7424836601307191, "eval_overall_precision": 0.6794258373205742, "eval_overall_recall": 0.8184438040345822, "eval_runtime": 0.3771, "eval_samples_per_second": 495.902, "eval_steps_per_second": 7.956, "step": 954 }, { "epoch": 10.0, "grad_norm": 1.1408404111862183, "learning_rate": 4.5e-05, "loss": 0.1834, "step": 1060 }, { "epoch": 10.0, "eval_LOCATION_f1": 0.7547169811320754, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.6818181818181818, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.6463414634146342, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5353535353535354, "eval_ORGANIZATION_recall": 0.8153846153846154, "eval_PERSON_f1": 0.8369230769230769, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.7771428571428571, "eval_PERSON_recall": 0.9066666666666666, "eval_QUANTITY_f1": 0.64, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.5714285714285714, "eval_QUANTITY_recall": 0.7272727272727273, "eval_TIME_f1": 0.7619047619047619, "eval_TIME_number": 28, "eval_TIME_precision": 0.6857142857142857, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.16224096715450287, "eval_overall_accuracy": 0.9456838021338506, "eval_overall_f1": 0.7557251908396947, "eval_overall_precision": 0.6765375854214123, "eval_overall_recall": 0.8559077809798271, "eval_runtime": 0.3807, "eval_samples_per_second": 491.168, "eval_steps_per_second": 7.88, "step": 1060 }, { "epoch": 11.0, "grad_norm": 2.472763776779175, "learning_rate": 4.4500000000000004e-05, "loss": 0.175, "step": 1166 }, { "epoch": 11.0, "eval_LOCATION_f1": 0.7439024390243902, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.6559139784946236, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.6710526315789473, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5862068965517241, "eval_ORGANIZATION_recall": 0.7846153846153846, "eval_PERSON_f1": 0.84375, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.7941176470588235, "eval_PERSON_recall": 0.9, "eval_QUANTITY_f1": 0.6493506493506495, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.5681818181818182, "eval_QUANTITY_recall": 0.7575757575757576, "eval_TIME_f1": 0.793103448275862, "eval_TIME_number": 28, "eval_TIME_precision": 0.7666666666666667, "eval_TIME_recall": 0.8214285714285714, "eval_loss": 0.14989984035491943, "eval_overall_accuracy": 0.950533462657614, "eval_overall_f1": 0.7652399481193255, "eval_overall_precision": 0.6957547169811321, "eval_overall_recall": 0.8501440922190202, "eval_runtime": 0.3778, "eval_samples_per_second": 494.952, "eval_steps_per_second": 7.94, "step": 1166 }, { "epoch": 12.0, "grad_norm": 2.2748563289642334, "learning_rate": 4.4000000000000006e-05, "loss": 0.1637, "step": 1272 }, { "epoch": 12.0, "eval_LOCATION_f1": 0.7625, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.6853932584269663, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.6624203821656051, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5652173913043478, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8473520249221183, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.7953216374269005, "eval_PERSON_recall": 0.9066666666666666, "eval_QUANTITY_f1": 0.6933333333333334, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6190476190476191, "eval_QUANTITY_recall": 0.7878787878787878, "eval_TIME_f1": 0.806451612903226, "eval_TIME_number": 28, "eval_TIME_precision": 0.7352941176470589, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.15028567612171173, "eval_overall_accuracy": 0.950533462657614, "eval_overall_f1": 0.7741935483870966, "eval_overall_precision": 0.7009345794392523, "eval_overall_recall": 0.8645533141210374, "eval_runtime": 0.3786, "eval_samples_per_second": 493.887, "eval_steps_per_second": 7.923, "step": 1272 }, { "epoch": 13.0, "grad_norm": 1.009729266166687, "learning_rate": 4.35e-05, "loss": 0.1569, "step": 1378 }, { "epoch": 13.0, "eval_LOCATION_f1": 0.7662337662337662, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7108433734939759, "eval_LOCATION_recall": 0.8309859154929577, "eval_ORGANIZATION_f1": 0.6583850931677019, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5520833333333334, "eval_ORGANIZATION_recall": 0.8153846153846154, "eval_PERSON_f1": 0.8473520249221183, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.7953216374269005, "eval_PERSON_recall": 0.9066666666666666, "eval_QUANTITY_f1": 0.6849315068493151, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.625, "eval_QUANTITY_recall": 0.7575757575757576, "eval_TIME_f1": 0.8620689655172413, "eval_TIME_number": 28, "eval_TIME_precision": 0.8333333333333334, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.14378908276557922, "eval_overall_accuracy": 0.9515033947623667, "eval_overall_f1": 0.7770534550195566, "eval_overall_precision": 0.7095238095238096, "eval_overall_recall": 0.8587896253602305, "eval_runtime": 0.3744, "eval_samples_per_second": 499.42, "eval_steps_per_second": 8.012, "step": 1378 }, { "epoch": 14.0, "grad_norm": 2.2329134941101074, "learning_rate": 4.3e-05, "loss": 0.1531, "step": 1484 }, { "epoch": 14.0, "eval_LOCATION_f1": 0.7682119205298014, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.725, "eval_LOCATION_recall": 0.8169014084507042, "eval_ORGANIZATION_f1": 0.6624203821656051, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5652173913043478, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8500000000000001, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8, "eval_PERSON_recall": 0.9066666666666666, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8474576271186439, "eval_TIME_number": 28, "eval_TIME_precision": 0.8064516129032258, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.14014975726604462, "eval_overall_accuracy": 0.9527158098933075, "eval_overall_f1": 0.7831800262812089, "eval_overall_precision": 0.7198067632850241, "eval_overall_recall": 0.8587896253602305, "eval_runtime": 0.3724, "eval_samples_per_second": 502.108, "eval_steps_per_second": 8.055, "step": 1484 }, { "epoch": 15.0, "grad_norm": 1.4270859956741333, "learning_rate": 4.25e-05, "loss": 0.1479, "step": 1590 }, { "epoch": 15.0, "eval_LOCATION_f1": 0.759493670886076, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.6896551724137931, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.6463414634146342, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5353535353535354, "eval_ORGANIZATION_recall": 0.8153846153846154, "eval_PERSON_f1": 0.8353658536585366, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.7696629213483146, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.7272727272727273, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6363636363636364, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8333333333333334, "eval_TIME_number": 28, "eval_TIME_precision": 0.78125, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.1550932377576828, "eval_overall_accuracy": 0.9488360814742968, "eval_overall_f1": 0.7700127064803051, "eval_overall_precision": 0.6886363636363636, "eval_overall_recall": 0.8731988472622478, "eval_runtime": 0.3738, "eval_samples_per_second": 500.322, "eval_steps_per_second": 8.027, "step": 1590 }, { "epoch": 16.0, "grad_norm": 1.5775573253631592, "learning_rate": 4.2e-05, "loss": 0.1429, "step": 1696 }, { "epoch": 16.0, "eval_LOCATION_f1": 0.7612903225806451, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7023809523809523, "eval_LOCATION_recall": 0.8309859154929577, "eval_ORGANIZATION_f1": 0.6538461538461537, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5604395604395604, "eval_ORGANIZATION_recall": 0.7846153846153846, "eval_PERSON_f1": 0.8589341692789968, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8106508875739645, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.7179487179487178, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6222222222222222, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.7999999999999999, "eval_TIME_number": 28, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.14502553641796112, "eval_overall_accuracy": 0.9515033947623667, "eval_overall_f1": 0.7786458333333334, "eval_overall_precision": 0.7102137767220903, "eval_overall_recall": 0.861671469740634, "eval_runtime": 0.3755, "eval_samples_per_second": 498.05, "eval_steps_per_second": 7.99, "step": 1696 }, { "epoch": 17.0, "grad_norm": 1.0638453960418701, "learning_rate": 4.15e-05, "loss": 0.1421, "step": 1802 }, { "epoch": 17.0, "eval_LOCATION_f1": 0.7631578947368421, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7160493827160493, "eval_LOCATION_recall": 0.8169014084507042, "eval_ORGANIZATION_f1": 0.6666666666666666, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5714285714285714, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8473520249221183, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.7953216374269005, "eval_PERSON_recall": 0.9066666666666666, "eval_QUANTITY_f1": 0.7567567567567567, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6829268292682927, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8333333333333334, "eval_TIME_number": 28, "eval_TIME_precision": 0.78125, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.13652156293392181, "eval_overall_accuracy": 0.9529582929194956, "eval_overall_f1": 0.7837483617300132, "eval_overall_precision": 0.71875, "eval_overall_recall": 0.861671469740634, "eval_runtime": 0.3757, "eval_samples_per_second": 497.696, "eval_steps_per_second": 7.984, "step": 1802 }, { "epoch": 18.0, "grad_norm": 1.2954357862472534, "learning_rate": 4.1e-05, "loss": 0.1385, "step": 1908 }, { "epoch": 18.0, "eval_LOCATION_f1": 0.7712418300653594, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7195121951219512, "eval_LOCATION_recall": 0.8309859154929577, "eval_ORGANIZATION_f1": 0.65, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5473684210526316, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8430769230769231, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.7828571428571428, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.7466666666666666, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6666666666666666, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8333333333333334, "eval_TIME_number": 28, "eval_TIME_precision": 0.78125, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.14212766289710999, "eval_overall_accuracy": 0.9517458777885548, "eval_overall_f1": 0.7787839586028461, "eval_overall_precision": 0.7065727699530516, "eval_overall_recall": 0.8674351585014409, "eval_runtime": 0.3862, "eval_samples_per_second": 484.163, "eval_steps_per_second": 7.767, "step": 1908 }, { "epoch": 19.0, "grad_norm": 1.54674232006073, "learning_rate": 4.05e-05, "loss": 0.1316, "step": 2014 }, { "epoch": 19.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.6790123456790124, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5670103092783505, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8643533123028392, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8203592814371258, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.7671232876712328, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.7, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8333333333333334, "eval_TIME_number": 28, "eval_TIME_precision": 0.78125, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.13803647458553314, "eval_overall_accuracy": 0.9536857419980601, "eval_overall_f1": 0.7953063885267275, "eval_overall_precision": 0.7261904761904762, "eval_overall_recall": 0.8789625360230547, "eval_runtime": 0.3784, "eval_samples_per_second": 494.235, "eval_steps_per_second": 7.929, "step": 2014 }, { "epoch": 20.0, "grad_norm": 1.5684103965759277, "learning_rate": 4e-05, "loss": 0.1305, "step": 2120 }, { "epoch": 20.0, "eval_LOCATION_f1": 0.7692307692307694, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7058823529411765, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.6797385620915033, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5909090909090909, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8643533123028392, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8203592814371258, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.7200000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6428571428571429, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.7999999999999999, "eval_TIME_number": 28, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13493511080741882, "eval_overall_accuracy": 0.9532007759456838, "eval_overall_f1": 0.7884362680683311, "eval_overall_precision": 0.7246376811594203, "eval_overall_recall": 0.8645533141210374, "eval_runtime": 0.3753, "eval_samples_per_second": 498.248, "eval_steps_per_second": 7.993, "step": 2120 }, { "epoch": 21.0, "grad_norm": 2.151451587677002, "learning_rate": 3.9500000000000005e-05, "loss": 0.127, "step": 2226 }, { "epoch": 21.0, "eval_LOCATION_f1": 0.7814569536423841, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.8309859154929577, "eval_ORGANIZATION_f1": 0.6797385620915033, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5909090909090909, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8607594936708859, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8192771084337349, "eval_PERSON_recall": 0.9066666666666666, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8813559322033899, "eval_TIME_number": 28, "eval_TIME_precision": 0.8387096774193549, "eval_TIME_recall": 0.9285714285714286, "eval_loss": 0.1296316385269165, "eval_overall_accuracy": 0.9558680892337537, "eval_overall_f1": 0.796812749003984, "eval_overall_precision": 0.7389162561576355, "eval_overall_recall": 0.8645533141210374, "eval_runtime": 0.3748, "eval_samples_per_second": 498.997, "eval_steps_per_second": 8.005, "step": 2226 }, { "epoch": 22.0, "grad_norm": 1.7709009647369385, "learning_rate": 3.9000000000000006e-05, "loss": 0.1256, "step": 2332 }, { "epoch": 22.0, "eval_LOCATION_f1": 0.7721518987341772, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7011494252873564, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.6586826347305389, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5392156862745098, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.845679012345679, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.7873563218390804, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.675, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.574468085106383, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.7868852459016394, "eval_TIME_number": 28, "eval_TIME_precision": 0.7272727272727273, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.15909084677696228, "eval_overall_accuracy": 0.9466537342386033, "eval_overall_f1": 0.769620253164557, "eval_overall_precision": 0.6862302483069977, "eval_overall_recall": 0.8760806916426513, "eval_runtime": 0.3738, "eval_samples_per_second": 500.205, "eval_steps_per_second": 8.025, "step": 2332 }, { "epoch": 23.0, "grad_norm": 1.6471892595291138, "learning_rate": 3.85e-05, "loss": 0.1241, "step": 2438 }, { "epoch": 23.0, "eval_LOCATION_f1": 0.7866666666666666, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7468354430379747, "eval_LOCATION_recall": 0.8309859154929577, "eval_ORGANIZATION_f1": 0.6666666666666667, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5567010309278351, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.8698412698412699, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8303030303030303, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.896551724137931, "eval_TIME_number": 28, "eval_TIME_precision": 0.8666666666666667, "eval_TIME_recall": 0.9285714285714286, "eval_loss": 0.13220064342021942, "eval_overall_accuracy": 0.954898157129001, "eval_overall_f1": 0.7984189723320159, "eval_overall_precision": 0.7354368932038835, "eval_overall_recall": 0.8731988472622478, "eval_runtime": 0.376, "eval_samples_per_second": 497.343, "eval_steps_per_second": 7.979, "step": 2438 }, { "epoch": 24.0, "grad_norm": 2.134507656097412, "learning_rate": 3.8e-05, "loss": 0.1204, "step": 2544 }, { "epoch": 24.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.6751592356687899, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5760869565217391, "eval_ORGANIZATION_recall": 0.8153846153846154, "eval_PERSON_f1": 0.8589341692789968, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8106508875739645, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8333333333333334, "eval_TIME_number": 28, "eval_TIME_precision": 0.78125, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.13271385431289673, "eval_overall_accuracy": 0.9544131910766246, "eval_overall_f1": 0.7905759162303665, "eval_overall_precision": 0.7242206235011991, "eval_overall_recall": 0.8703170028818443, "eval_runtime": 0.3795, "eval_samples_per_second": 492.774, "eval_steps_per_second": 7.905, "step": 2544 }, { "epoch": 25.0, "grad_norm": 3.391195774078369, "learning_rate": 3.7500000000000003e-05, "loss": 0.1202, "step": 2650 }, { "epoch": 25.0, "eval_LOCATION_f1": 0.7843137254901961, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7317073170731707, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.6666666666666667, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5567010309278351, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.85625, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8058823529411765, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.6835443037974683, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.5869565217391305, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.7999999999999999, "eval_TIME_number": 28, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.14239878952503204, "eval_overall_accuracy": 0.9502909796314258, "eval_overall_f1": 0.7803617571059431, "eval_overall_precision": 0.7072599531615925, "eval_overall_recall": 0.8703170028818443, "eval_runtime": 0.3766, "eval_samples_per_second": 496.546, "eval_steps_per_second": 7.966, "step": 2650 }, { "epoch": 26.0, "grad_norm": 1.319371223449707, "learning_rate": 3.7e-05, "loss": 0.1162, "step": 2756 }, { "epoch": 26.0, "eval_LOCATION_f1": 0.7919463087248323, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7564102564102564, "eval_LOCATION_recall": 0.8309859154929577, "eval_ORGANIZATION_f1": 0.6878980891719746, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5869565217391305, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.8598130841121495, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8070175438596491, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7027027027027027, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6341463414634146, "eval_QUANTITY_recall": 0.7878787878787878, "eval_TIME_f1": 0.7999999999999999, "eval_TIME_number": 28, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1371098905801773, "eval_overall_accuracy": 0.9536857419980601, "eval_overall_f1": 0.7910643889618922, "eval_overall_precision": 0.7270531400966184, "eval_overall_recall": 0.8674351585014409, "eval_runtime": 0.3768, "eval_samples_per_second": 496.253, "eval_steps_per_second": 7.961, "step": 2756 }, { "epoch": 27.0, "grad_norm": 1.0089757442474365, "learning_rate": 3.65e-05, "loss": 0.113, "step": 2862 }, { "epoch": 27.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.675, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5684210526315789, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.8535825545171339, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8011695906432749, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.7200000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6428571428571429, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.7999999999999999, "eval_TIME_number": 28, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1443266123533249, "eval_overall_accuracy": 0.9515033947623667, "eval_overall_f1": 0.7833981841763942, "eval_overall_precision": 0.7122641509433962, "eval_overall_recall": 0.8703170028818443, "eval_runtime": 0.3774, "eval_samples_per_second": 495.464, "eval_steps_per_second": 7.949, "step": 2862 }, { "epoch": 28.0, "grad_norm": 2.329519748687744, "learning_rate": 3.6e-05, "loss": 0.1121, "step": 2968 }, { "epoch": 28.0, "eval_LOCATION_f1": 0.7894736842105262, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7407407407407407, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5894736842105263, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8625, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8117647058823529, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.6933333333333334, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6190476190476191, "eval_QUANTITY_recall": 0.7878787878787878, "eval_TIME_f1": 0.8333333333333334, "eval_TIME_number": 28, "eval_TIME_precision": 0.78125, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.13519856333732605, "eval_overall_accuracy": 0.9544131910766246, "eval_overall_f1": 0.7953063885267275, "eval_overall_precision": 0.7261904761904762, "eval_overall_recall": 0.8789625360230547, "eval_runtime": 0.3777, "eval_samples_per_second": 495.05, "eval_steps_per_second": 7.942, "step": 2968 }, { "epoch": 29.0, "grad_norm": 0.6808164715766907, "learning_rate": 3.55e-05, "loss": 0.1127, "step": 3074 }, { "epoch": 29.0, "eval_LOCATION_f1": 0.8026315789473684, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7530864197530864, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.6748466257668712, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5612244897959183, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8652037617554857, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8165680473372781, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7027027027027027, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6341463414634146, "eval_QUANTITY_recall": 0.7878787878787878, "eval_TIME_f1": 0.8333333333333334, "eval_TIME_number": 28, "eval_TIME_precision": 0.78125, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.13588234782218933, "eval_overall_accuracy": 0.9529582929194956, "eval_overall_f1": 0.7942708333333333, "eval_overall_precision": 0.7244655581947743, "eval_overall_recall": 0.8789625360230547, "eval_runtime": 0.3768, "eval_samples_per_second": 496.222, "eval_steps_per_second": 7.961, "step": 3074 }, { "epoch": 30.0, "grad_norm": 1.221112847328186, "learning_rate": 3.5e-05, "loss": 0.1101, "step": 3180 }, { "epoch": 30.0, "eval_LOCATION_f1": 0.7843137254901961, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7317073170731707, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.6666666666666667, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5567010309278351, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.8571428571428572, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8023255813953488, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8333333333333334, "eval_TIME_number": 28, "eval_TIME_precision": 0.78125, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.1349276304244995, "eval_overall_accuracy": 0.953443258971872, "eval_overall_f1": 0.7896103896103897, "eval_overall_precision": 0.7186761229314421, "eval_overall_recall": 0.8760806916426513, "eval_runtime": 0.377, "eval_samples_per_second": 495.999, "eval_steps_per_second": 7.957, "step": 3180 }, { "epoch": 31.0, "grad_norm": 0.5136931538581848, "learning_rate": 3.45e-05, "loss": 0.1085, "step": 3286 }, { "epoch": 31.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.6792452830188681, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.574468085106383, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.8652037617554857, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8165680473372781, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7123287671232875, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.65, "eval_QUANTITY_recall": 0.7878787878787878, "eval_TIME_f1": 0.8333333333333334, "eval_TIME_number": 28, "eval_TIME_precision": 0.78125, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.1328558474779129, "eval_overall_accuracy": 0.9544131910766246, "eval_overall_f1": 0.7911227154046998, "eval_overall_precision": 0.7231503579952268, "eval_overall_recall": 0.8731988472622478, "eval_runtime": 0.3748, "eval_samples_per_second": 498.917, "eval_steps_per_second": 8.004, "step": 3286 }, { "epoch": 32.0, "grad_norm": 2.1277899742126465, "learning_rate": 3.4000000000000007e-05, "loss": 0.1055, "step": 3392 }, { "epoch": 32.0, "eval_LOCATION_f1": 0.7770700636942676, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7093023255813954, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.7, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5894736842105263, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8625, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8117647058823529, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8333333333333334, "eval_TIME_number": 28, "eval_TIME_precision": 0.78125, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.14278312027454376, "eval_overall_accuracy": 0.9536857419980601, "eval_overall_f1": 0.7963683527885862, "eval_overall_precision": 0.7240566037735849, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3755, "eval_samples_per_second": 498.064, "eval_steps_per_second": 7.99, "step": 3392 }, { "epoch": 33.0, "grad_norm": 1.8276546001434326, "learning_rate": 3.35e-05, "loss": 0.1058, "step": 3498 }, { "epoch": 33.0, "eval_LOCATION_f1": 0.794701986754967, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.75, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.6962025316455697, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5913978494623656, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8769716088328076, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8323353293413174, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7500000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6923076923076923, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8333333333333334, "eval_TIME_number": 28, "eval_TIME_precision": 0.78125, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.1318831443786621, "eval_overall_accuracy": 0.9568380213385063, "eval_overall_f1": 0.8073878627968336, "eval_overall_precision": 0.7445255474452555, "eval_overall_recall": 0.8818443804034583, "eval_runtime": 0.3744, "eval_samples_per_second": 499.447, "eval_steps_per_second": 8.013, "step": 3498 }, { "epoch": 34.0, "grad_norm": 1.7405362129211426, "learning_rate": 3.3e-05, "loss": 0.1038, "step": 3604 }, { "epoch": 34.0, "eval_LOCATION_f1": 0.7792207792207793, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7228915662650602, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.708860759493671, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6021505376344086, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8917197452229298, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8536585365853658, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.7999999999999999, "eval_TIME_number": 28, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13387276232242584, "eval_overall_accuracy": 0.9565955383123181, "eval_overall_f1": 0.8089591567852438, "eval_overall_precision": 0.7451456310679612, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3759, "eval_samples_per_second": 497.447, "eval_steps_per_second": 7.98, "step": 3604 }, { "epoch": 35.0, "grad_norm": 0.9519175887107849, "learning_rate": 3.2500000000000004e-05, "loss": 0.1055, "step": 3710 }, { "epoch": 35.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.713375796178344, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6086956521739131, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8769716088328076, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8323353293413174, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.7999999999999999, "eval_TIME_number": 28, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1373051106929779, "eval_overall_accuracy": 0.9553831231813773, "eval_overall_f1": 0.8020969855832241, "eval_overall_precision": 0.7355769230769231, "eval_overall_recall": 0.8818443804034583, "eval_runtime": 0.3734, "eval_samples_per_second": 500.858, "eval_steps_per_second": 8.035, "step": 3710 }, { "epoch": 36.0, "grad_norm": 1.7379624843597412, "learning_rate": 3.2000000000000005e-05, "loss": 0.1012, "step": 3816 }, { "epoch": 36.0, "eval_LOCATION_f1": 0.7870967741935485, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7261904761904762, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.691358024691358, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5773195876288659, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8769716088328076, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8323353293413174, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.6923076923076923, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8333333333333334, "eval_TIME_number": 28, "eval_TIME_precision": 0.78125, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.14059388637542725, "eval_overall_accuracy": 0.9536857419980601, "eval_overall_f1": 0.7979274611398965, "eval_overall_precision": 0.7247058823529412, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.374, "eval_samples_per_second": 500.026, "eval_steps_per_second": 8.022, "step": 3816 }, { "epoch": 37.0, "grad_norm": 1.2913354635238647, "learning_rate": 3.15e-05, "loss": 0.0982, "step": 3922 }, { "epoch": 37.0, "eval_LOCATION_f1": 0.782051282051282, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7176470588235294, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.6875, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5789473684210527, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8832807570977917, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8383233532934131, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7105263157894738, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.627906976744186, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.7868852459016394, "eval_TIME_number": 28, "eval_TIME_precision": 0.7272727272727273, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.14024104177951813, "eval_overall_accuracy": 0.953443258971872, "eval_overall_f1": 0.7974025974025974, "eval_overall_precision": 0.7257683215130024, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3755, "eval_samples_per_second": 498.059, "eval_steps_per_second": 7.99, "step": 3922 }, { "epoch": 38.0, "grad_norm": 0.7977638244628906, "learning_rate": 3.1e-05, "loss": 0.0982, "step": 4028 }, { "epoch": 38.0, "eval_LOCATION_f1": 0.782051282051282, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7176470588235294, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.7051282051282052, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6043956043956044, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.888888888888889, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8484848484848485, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7200000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6428571428571429, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.7868852459016394, "eval_TIME_number": 28, "eval_TIME_precision": 0.7272727272727273, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13799616694450378, "eval_overall_accuracy": 0.9553831231813773, "eval_overall_f1": 0.8047182175622541, "eval_overall_precision": 0.7379807692307693, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3743, "eval_samples_per_second": 499.559, "eval_steps_per_second": 8.014, "step": 4028 }, { "epoch": 39.0, "grad_norm": 1.380632758140564, "learning_rate": 3.05e-05, "loss": 0.0976, "step": 4134 }, { "epoch": 39.0, "eval_LOCATION_f1": 0.8, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7380952380952381, "eval_LOCATION_recall": 0.8732394366197183, "eval_ORGANIZATION_f1": 0.6956521739130436, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5833333333333334, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8742138364779874, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8273809523809523, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7123287671232875, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.65, "eval_QUANTITY_recall": 0.7878787878787878, "eval_TIME_f1": 0.7868852459016394, "eval_TIME_number": 28, "eval_TIME_precision": 0.7272727272727273, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13894657790660858, "eval_overall_accuracy": 0.9544131910766246, "eval_overall_f1": 0.7994791666666666, "eval_overall_precision": 0.7292161520190024, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3767, "eval_samples_per_second": 496.398, "eval_steps_per_second": 7.964, "step": 4134 }, { "epoch": 40.0, "grad_norm": 1.3057403564453125, "learning_rate": 3e-05, "loss": 0.0987, "step": 4240 }, { "epoch": 40.0, "eval_LOCATION_f1": 0.8, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7380952380952381, "eval_LOCATION_recall": 0.8732394366197183, "eval_ORGANIZATION_f1": 0.7225806451612903, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6222222222222222, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8734177215189873, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8313253012048193, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8620689655172413, "eval_TIME_number": 28, "eval_TIME_precision": 0.8333333333333334, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.1264629065990448, "eval_overall_accuracy": 0.9590203685741998, "eval_overall_f1": 0.8137384412153236, "eval_overall_precision": 0.751219512195122, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.3763, "eval_samples_per_second": 496.897, "eval_steps_per_second": 7.972, "step": 4240 }, { "epoch": 41.0, "grad_norm": 0.4240831434726715, "learning_rate": 2.95e-05, "loss": 0.0975, "step": 4346 }, { "epoch": 41.0, "eval_LOCATION_f1": 0.7922077922077921, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7349397590361446, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.713375796178344, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6086956521739131, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8832807570977917, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8383233532934131, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7200000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6428571428571429, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8333333333333334, "eval_TIME_number": 28, "eval_TIME_precision": 0.78125, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.13584652543067932, "eval_overall_accuracy": 0.95635305528613, "eval_overall_f1": 0.8099606815203145, "eval_overall_precision": 0.7427884615384616, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.3745, "eval_samples_per_second": 499.345, "eval_steps_per_second": 8.011, "step": 4346 }, { "epoch": 42.0, "grad_norm": 1.1680493354797363, "learning_rate": 2.9e-05, "loss": 0.0963, "step": 4452 }, { "epoch": 42.0, "eval_LOCATION_f1": 0.7843137254901961, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7317073170731707, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.717948717948718, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6153846153846154, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.888888888888889, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8484848484848485, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7027027027027027, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6341463414634146, "eval_QUANTITY_recall": 0.7878787878787878, "eval_TIME_f1": 0.8474576271186439, "eval_TIME_number": 28, "eval_TIME_precision": 0.8064516129032258, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.1361624002456665, "eval_overall_accuracy": 0.9565955383123181, "eval_overall_f1": 0.8110964332892999, "eval_overall_precision": 0.748780487804878, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3755, "eval_samples_per_second": 498.014, "eval_steps_per_second": 7.99, "step": 4452 }, { "epoch": 43.0, "grad_norm": 0.6009380221366882, "learning_rate": 2.8499999999999998e-05, "loss": 0.0956, "step": 4558 }, { "epoch": 43.0, "eval_LOCATION_f1": 0.8051948051948051, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7469879518072289, "eval_LOCATION_recall": 0.8732394366197183, "eval_ORGANIZATION_f1": 0.728476821192053, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6395348837209303, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8853503184713376, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8475609756097561, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7222222222222221, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6666666666666666, "eval_QUANTITY_recall": 0.7878787878787878, "eval_TIME_f1": 0.8474576271186439, "eval_TIME_number": 28, "eval_TIME_precision": 0.8064516129032258, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.1262454241514206, "eval_overall_accuracy": 0.9597478176527643, "eval_overall_f1": 0.8186666666666667, "eval_overall_precision": 0.7617866004962779, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3745, "eval_samples_per_second": 499.297, "eval_steps_per_second": 8.01, "step": 4558 }, { "epoch": 44.0, "grad_norm": 1.364172339439392, "learning_rate": 2.8000000000000003e-05, "loss": 0.0957, "step": 4664 }, { "epoch": 44.0, "eval_LOCATION_f1": 0.7922077922077921, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7349397590361446, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.728476821192053, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6395348837209303, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8917197452229298, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8536585365853658, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8474576271186439, "eval_TIME_number": 28, "eval_TIME_precision": 0.8064516129032258, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.12773707509040833, "eval_overall_accuracy": 0.9599903006789525, "eval_overall_f1": 0.8202396804260986, "eval_overall_precision": 0.7623762376237624, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.3737, "eval_samples_per_second": 500.403, "eval_steps_per_second": 8.028, "step": 4664 }, { "epoch": 45.0, "grad_norm": 8.042428016662598, "learning_rate": 2.7500000000000004e-05, "loss": 0.0931, "step": 4770 }, { "epoch": 45.0, "eval_LOCATION_f1": 0.7948717948717949, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7294117647058823, "eval_LOCATION_recall": 0.8732394366197183, "eval_ORGANIZATION_f1": 0.7272727272727273, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6292134831460674, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8917197452229298, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8536585365853658, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8474576271186439, "eval_TIME_number": 28, "eval_TIME_precision": 0.8064516129032258, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.12983648478984833, "eval_overall_accuracy": 0.9595053346265762, "eval_overall_f1": 0.8201058201058202, "eval_overall_precision": 0.7579462102689487, "eval_overall_recall": 0.8933717579250721, "eval_runtime": 0.3781, "eval_samples_per_second": 494.639, "eval_steps_per_second": 7.935, "step": 4770 }, { "epoch": 46.0, "grad_norm": 1.4171274900436401, "learning_rate": 2.7000000000000002e-05, "loss": 0.0944, "step": 4876 }, { "epoch": 46.0, "eval_LOCATION_f1": 0.794701986754967, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.75, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7142857142857143, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6179775280898876, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.888178913738019, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.852760736196319, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7887323943661972, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.7368421052631579, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8620689655172413, "eval_TIME_number": 28, "eval_TIME_precision": 0.8333333333333334, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.12546101212501526, "eval_overall_accuracy": 0.9597478176527643, "eval_overall_f1": 0.821954484605087, "eval_overall_precision": 0.7675, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3758, "eval_samples_per_second": 497.637, "eval_steps_per_second": 7.983, "step": 4876 }, { "epoch": 47.0, "grad_norm": 1.7135871648788452, "learning_rate": 2.6500000000000004e-05, "loss": 0.0908, "step": 4982 }, { "epoch": 47.0, "eval_LOCATION_f1": 0.7770700636942676, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7093023255813954, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.7225806451612903, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6222222222222222, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8917197452229298, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8536585365853658, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13818912208080292, "eval_overall_accuracy": 0.9570805043646945, "eval_overall_f1": 0.8126649076517151, "eval_overall_precision": 0.7493917274939172, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.3741, "eval_samples_per_second": 499.91, "eval_steps_per_second": 8.02, "step": 4982 }, { "epoch": 48.0, "grad_norm": 3.0958287715911865, "learning_rate": 2.6000000000000002e-05, "loss": 0.0926, "step": 5088 }, { "epoch": 48.0, "eval_LOCATION_f1": 0.782051282051282, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7176470588235294, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.717948717948718, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6153846153846154, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8917197452229298, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8536585365853658, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8474576271186439, "eval_TIME_number": 28, "eval_TIME_precision": 0.8064516129032258, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.13272243738174438, "eval_overall_accuracy": 0.9587778855480117, "eval_overall_f1": 0.8153034300791556, "eval_overall_precision": 0.7518248175182481, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.3743, "eval_samples_per_second": 499.613, "eval_steps_per_second": 8.015, "step": 5088 }, { "epoch": 49.0, "grad_norm": 0.8228998780250549, "learning_rate": 2.5500000000000003e-05, "loss": 0.0906, "step": 5194 }, { "epoch": 49.0, "eval_LOCATION_f1": 0.7848101265822784, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7126436781609196, "eval_LOCATION_recall": 0.8732394366197183, "eval_ORGANIZATION_f1": 0.717948717948718, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6153846153846154, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8980891719745222, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8597560975609756, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7200000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6428571428571429, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.7666666666666666, "eval_TIME_number": 28, "eval_TIME_precision": 0.71875, "eval_TIME_recall": 0.8214285714285714, "eval_loss": 0.13896866142749786, "eval_overall_accuracy": 0.9568380213385063, "eval_overall_f1": 0.8099606815203145, "eval_overall_precision": 0.7427884615384616, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.3751, "eval_samples_per_second": 498.49, "eval_steps_per_second": 7.997, "step": 5194 }, { "epoch": 50.0, "grad_norm": 1.645812749862671, "learning_rate": 2.5e-05, "loss": 0.0875, "step": 5300 }, { "epoch": 50.0, "eval_LOCATION_f1": 0.782051282051282, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7176470588235294, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.7225806451612903, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6222222222222222, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8924050632911392, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8493975903614458, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1355874240398407, "eval_overall_accuracy": 0.957807953443259, "eval_overall_f1": 0.8142292490118578, "eval_overall_precision": 0.75, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.3761, "eval_samples_per_second": 497.165, "eval_steps_per_second": 7.976, "step": 5300 }, { "epoch": 51.0, "grad_norm": 1.3679982423782349, "learning_rate": 2.45e-05, "loss": 0.0885, "step": 5406 }, { "epoch": 51.0, "eval_LOCATION_f1": 0.7898089171974522, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7209302325581395, "eval_LOCATION_recall": 0.8732394366197183, "eval_ORGANIZATION_f1": 0.717948717948718, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6153846153846154, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9009584664536742, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8650306748466258, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13315366208553314, "eval_overall_accuracy": 0.9585354025218235, "eval_overall_f1": 0.8168642951251647, "eval_overall_precision": 0.7524271844660194, "eval_overall_recall": 0.8933717579250721, "eval_runtime": 0.375, "eval_samples_per_second": 498.669, "eval_steps_per_second": 8.0, "step": 5406 }, { "epoch": 52.0, "grad_norm": 0.8684185743331909, "learning_rate": 2.4e-05, "loss": 0.0886, "step": 5512 }, { "epoch": 52.0, "eval_LOCATION_f1": 0.7922077922077921, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7349397590361446, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.7096774193548387, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6111111111111112, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8980891719745222, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8597560975609756, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7200000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6428571428571429, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13661393523216248, "eval_overall_accuracy": 0.9575654704170709, "eval_overall_f1": 0.8137384412153236, "eval_overall_precision": 0.751219512195122, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.3753, "eval_samples_per_second": 498.274, "eval_steps_per_second": 7.994, "step": 5512 }, { "epoch": 53.0, "grad_norm": 1.742406964302063, "learning_rate": 2.35e-05, "loss": 0.0875, "step": 5618 }, { "epoch": 53.0, "eval_LOCATION_f1": 0.782051282051282, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7176470588235294, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.675, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5684210526315789, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.877742946708464, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8284023668639053, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7200000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6428571428571429, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.7796610169491526, "eval_TIME_number": 28, "eval_TIME_precision": 0.7419354838709677, "eval_TIME_recall": 0.8214285714285714, "eval_loss": 0.1452857404947281, "eval_overall_accuracy": 0.9541707080504365, "eval_overall_f1": 0.7932379713914174, "eval_overall_precision": 0.7227488151658767, "eval_overall_recall": 0.8789625360230547, "eval_runtime": 0.376, "eval_samples_per_second": 497.377, "eval_steps_per_second": 7.979, "step": 5618 }, { "epoch": 54.0, "grad_norm": 0.6719706654548645, "learning_rate": 2.3000000000000003e-05, "loss": 0.0821, "step": 5724 }, { "epoch": 54.0, "eval_LOCATION_f1": 0.7792207792207793, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7228915662650602, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7012987012987013, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6067415730337079, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.9067524115755627, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8757763975155279, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8474576271186439, "eval_TIME_number": 28, "eval_TIME_precision": 0.8064516129032258, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.1333644837141037, "eval_overall_accuracy": 0.9587778855480117, "eval_overall_f1": 0.8175765645805593, "eval_overall_precision": 0.7599009900990099, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3751, "eval_samples_per_second": 498.541, "eval_steps_per_second": 7.998, "step": 5724 }, { "epoch": 55.0, "grad_norm": 1.031419277191162, "learning_rate": 2.25e-05, "loss": 0.0839, "step": 5830 }, { "epoch": 55.0, "eval_LOCATION_f1": 0.7792207792207793, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7228915662650602, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7012987012987013, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6067415730337079, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.9015873015873016, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8606060606060606, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7200000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6428571428571429, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13826771080493927, "eval_overall_accuracy": 0.9582929194956353, "eval_overall_f1": 0.8110964332892999, "eval_overall_precision": 0.748780487804878, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3749, "eval_samples_per_second": 498.795, "eval_steps_per_second": 8.002, "step": 5830 }, { "epoch": 56.0, "grad_norm": 1.8279228210449219, "learning_rate": 2.2000000000000003e-05, "loss": 0.0886, "step": 5936 }, { "epoch": 56.0, "eval_LOCATION_f1": 0.7843137254901961, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7317073170731707, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7142857142857143, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6179775280898876, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8853503184713376, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8475609756097561, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7671232876712328, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.7, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8474576271186439, "eval_TIME_number": 28, "eval_TIME_precision": 0.8064516129032258, "eval_TIME_recall": 0.8928571428571429, "eval_loss": 0.13046370446681976, "eval_overall_accuracy": 0.9590203685741998, "eval_overall_f1": 0.8154050464807437, "eval_overall_precision": 0.7561576354679803, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3753, "eval_samples_per_second": 498.255, "eval_steps_per_second": 7.993, "step": 5936 }, { "epoch": 57.0, "grad_norm": 2.0078160762786865, "learning_rate": 2.15e-05, "loss": 0.0838, "step": 6042 }, { "epoch": 57.0, "eval_LOCATION_f1": 0.7843137254901961, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7317073170731707, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7189542483660131, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.625, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8952380952380952, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8545454545454545, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13407622277736664, "eval_overall_accuracy": 0.9570805043646945, "eval_overall_f1": 0.8143236074270557, "eval_overall_precision": 0.7542997542997543, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3763, "eval_samples_per_second": 496.984, "eval_steps_per_second": 7.973, "step": 6042 }, { "epoch": 58.0, "grad_norm": 1.1605350971221924, "learning_rate": 2.1e-05, "loss": 0.0846, "step": 6148 }, { "epoch": 58.0, "eval_LOCATION_f1": 0.7843137254901961, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7317073170731707, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7248322147651007, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6428571428571429, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.888888888888889, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8484848484848485, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7777777777777778, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.717948717948718, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.12788650393486023, "eval_overall_accuracy": 0.9597478176527643, "eval_overall_f1": 0.8181818181818181, "eval_overall_precision": 0.7630922693266833, "eval_overall_recall": 0.8818443804034583, "eval_runtime": 0.3761, "eval_samples_per_second": 497.243, "eval_steps_per_second": 7.977, "step": 6148 }, { "epoch": 59.0, "grad_norm": 1.6264705657958984, "learning_rate": 2.05e-05, "loss": 0.0832, "step": 6254 }, { "epoch": 59.0, "eval_LOCATION_f1": 0.7843137254901961, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7317073170731707, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7189542483660131, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.625, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8952380952380952, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8545454545454545, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7200000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6428571428571429, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13487118482589722, "eval_overall_accuracy": 0.9582929194956353, "eval_overall_f1": 0.8132450331125827, "eval_overall_precision": 0.7524509803921569, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3749, "eval_samples_per_second": 498.792, "eval_steps_per_second": 8.002, "step": 6254 }, { "epoch": 60.0, "grad_norm": 1.4992756843566895, "learning_rate": 2e-05, "loss": 0.0837, "step": 6360 }, { "epoch": 60.0, "eval_LOCATION_f1": 0.7894736842105262, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7407407407407407, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.738255033557047, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6547619047619048, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8860759493670886, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8433734939759037, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7671232876712328, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.7, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.12829264998435974, "eval_overall_accuracy": 0.9592628516003879, "eval_overall_f1": 0.8197596795727637, "eval_overall_precision": 0.763681592039801, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3751, "eval_samples_per_second": 498.47, "eval_steps_per_second": 7.997, "step": 6360 }, { "epoch": 61.0, "grad_norm": 1.2616682052612305, "learning_rate": 1.9500000000000003e-05, "loss": 0.0819, "step": 6466 }, { "epoch": 61.0, "eval_LOCATION_f1": 0.7712418300653594, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7195121951219512, "eval_LOCATION_recall": 0.8309859154929577, "eval_ORGANIZATION_f1": 0.7236842105263158, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.632183908045977, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8860759493670886, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8433734939759037, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7671232876712328, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.7, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13355708122253418, "eval_overall_accuracy": 0.9590203685741998, "eval_overall_f1": 0.8127490039840637, "eval_overall_precision": 0.7536945812807881, "eval_overall_recall": 0.8818443804034583, "eval_runtime": 0.3721, "eval_samples_per_second": 502.563, "eval_steps_per_second": 8.063, "step": 6466 }, { "epoch": 62.0, "grad_norm": 1.5365849733352661, "learning_rate": 1.9e-05, "loss": 0.0811, "step": 6572 }, { "epoch": 62.0, "eval_LOCATION_f1": 0.7712418300653594, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7195121951219512, "eval_LOCATION_recall": 0.8309859154929577, "eval_ORGANIZATION_f1": 0.7189542483660131, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.625, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9073482428115016, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8711656441717791, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13568484783172607, "eval_overall_accuracy": 0.9580504364694471, "eval_overall_f1": 0.8164893617021276, "eval_overall_precision": 0.7580246913580246, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3761, "eval_samples_per_second": 497.199, "eval_steps_per_second": 7.976, "step": 6572 }, { "epoch": 63.0, "grad_norm": 1.552751064300537, "learning_rate": 1.85e-05, "loss": 0.0796, "step": 6678 }, { "epoch": 63.0, "eval_LOCATION_f1": 0.7792207792207793, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7228915662650602, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7417218543046358, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6511627906976745, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9038461538461539, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8703703703703703, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7777777777777778, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.717948717948718, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.12987476587295532, "eval_overall_accuracy": 0.9602327837051406, "eval_overall_f1": 0.8273092369477911, "eval_overall_precision": 0.7725, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.3729, "eval_samples_per_second": 501.52, "eval_steps_per_second": 8.046, "step": 6678 }, { "epoch": 64.0, "grad_norm": 0.5724718570709229, "learning_rate": 1.8e-05, "loss": 0.0779, "step": 6784 }, { "epoch": 64.0, "eval_LOCATION_f1": 0.7843137254901961, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7317073170731707, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7333333333333334, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6470588235294118, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8924050632911392, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8493975903614458, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7500000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6923076923076923, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13085511326789856, "eval_overall_accuracy": 0.9595053346265762, "eval_overall_f1": 0.8186666666666667, "eval_overall_precision": 0.7617866004962779, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3731, "eval_samples_per_second": 501.257, "eval_steps_per_second": 8.042, "step": 6784 }, { "epoch": 65.0, "grad_norm": 1.46877920627594, "learning_rate": 1.75e-05, "loss": 0.0793, "step": 6890 }, { "epoch": 65.0, "eval_LOCATION_f1": 0.7894736842105262, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7407407407407407, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7225806451612903, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6222222222222222, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9073482428115016, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8711656441717791, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13223972916603088, "eval_overall_accuracy": 0.9587778855480117, "eval_overall_f1": 0.8207171314741034, "eval_overall_precision": 0.7610837438423645, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.3755, "eval_samples_per_second": 498.048, "eval_steps_per_second": 7.99, "step": 6890 }, { "epoch": 66.0, "grad_norm": 1.292722225189209, "learning_rate": 1.7000000000000003e-05, "loss": 0.0779, "step": 6996 }, { "epoch": 66.0, "eval_LOCATION_f1": 0.7870967741935485, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7261904761904762, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.7051282051282052, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6043956043956044, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8987341772151898, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8554216867469879, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.14010098576545715, "eval_overall_accuracy": 0.9575654704170709, "eval_overall_f1": 0.8131578947368422, "eval_overall_precision": 0.7481840193704601, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.3743, "eval_samples_per_second": 499.554, "eval_steps_per_second": 8.014, "step": 6996 }, { "epoch": 67.0, "grad_norm": 1.3288254737854004, "learning_rate": 1.65e-05, "loss": 0.0786, "step": 7102 }, { "epoch": 67.0, "eval_LOCATION_f1": 0.7792207792207793, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7228915662650602, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.738255033557047, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6547619047619048, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9102564102564102, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8765432098765432, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7500000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6923076923076923, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13140466809272766, "eval_overall_accuracy": 0.9604752667313288, "eval_overall_f1": 0.8257372654155496, "eval_overall_precision": 0.7719298245614035, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.375, "eval_samples_per_second": 498.682, "eval_steps_per_second": 8.0, "step": 7102 }, { "epoch": 68.0, "grad_norm": 1.3026788234710693, "learning_rate": 1.6000000000000003e-05, "loss": 0.0767, "step": 7208 }, { "epoch": 68.0, "eval_LOCATION_f1": 0.782051282051282, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7176470588235294, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.738255033557047, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6547619047619048, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9073482428115016, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8711656441717791, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.133492112159729, "eval_overall_accuracy": 0.9602327837051406, "eval_overall_f1": 0.8229027962716378, "eval_overall_precision": 0.7648514851485149, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.388, "eval_samples_per_second": 481.923, "eval_steps_per_second": 7.731, "step": 7208 }, { "epoch": 69.0, "grad_norm": 1.9889540672302246, "learning_rate": 1.55e-05, "loss": 0.0795, "step": 7314 }, { "epoch": 69.0, "eval_LOCATION_f1": 0.782051282051282, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7176470588235294, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.717948717948718, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6153846153846154, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8860759493670886, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8433734939759037, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7567567567567567, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6829268292682927, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13569195568561554, "eval_overall_accuracy": 0.957807953443259, "eval_overall_f1": 0.8131578947368422, "eval_overall_precision": 0.7481840193704601, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.3733, "eval_samples_per_second": 500.955, "eval_steps_per_second": 8.037, "step": 7314 }, { "epoch": 70.0, "grad_norm": 0.8371847867965698, "learning_rate": 1.5e-05, "loss": 0.0779, "step": 7420 }, { "epoch": 70.0, "eval_LOCATION_f1": 0.7792207792207793, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7228915662650602, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7333333333333334, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6470588235294118, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8860759493670886, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8433734939759037, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1332496553659439, "eval_overall_accuracy": 0.9592628516003879, "eval_overall_f1": 0.8149134487350199, "eval_overall_precision": 0.7574257425742574, "eval_overall_recall": 0.8818443804034583, "eval_runtime": 0.3763, "eval_samples_per_second": 496.984, "eval_steps_per_second": 7.973, "step": 7420 }, { "epoch": 71.0, "grad_norm": 2.211169958114624, "learning_rate": 1.45e-05, "loss": 0.0774, "step": 7526 }, { "epoch": 71.0, "eval_LOCATION_f1": 0.7770700636942676, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7093023255813954, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.713375796178344, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6086956521739131, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9044585987261147, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8658536585365854, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7200000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6428571428571429, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.7999999999999999, "eval_TIME_number": 28, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1430595964193344, "eval_overall_accuracy": 0.9568380213385063, "eval_overall_f1": 0.8125819134993448, "eval_overall_precision": 0.7451923076923077, "eval_overall_recall": 0.8933717579250721, "eval_runtime": 0.3768, "eval_samples_per_second": 496.324, "eval_steps_per_second": 7.962, "step": 7526 }, { "epoch": 72.0, "grad_norm": 1.1917625665664673, "learning_rate": 1.4000000000000001e-05, "loss": 0.0769, "step": 7632 }, { "epoch": 72.0, "eval_LOCATION_f1": 0.7792207792207793, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7228915662650602, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7432432432432433, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6626506024096386, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9009584664536742, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8650306748466258, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13343024253845215, "eval_overall_accuracy": 0.9602327837051406, "eval_overall_f1": 0.821954484605087, "eval_overall_precision": 0.7675, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3739, "eval_samples_per_second": 500.092, "eval_steps_per_second": 8.023, "step": 7632 }, { "epoch": 73.0, "grad_norm": 0.37685540318489075, "learning_rate": 1.3500000000000001e-05, "loss": 0.077, "step": 7738 }, { "epoch": 73.0, "eval_LOCATION_f1": 0.7922077922077921, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7349397590361446, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.738255033557047, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6547619047619048, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8910256410256411, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8580246913580247, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7777777777777778, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.717948717948718, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13034062087535858, "eval_overall_accuracy": 0.9609602327837051, "eval_overall_f1": 0.8241610738255033, "eval_overall_precision": 0.7713567839195979, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3772, "eval_samples_per_second": 495.809, "eval_steps_per_second": 7.954, "step": 7738 }, { "epoch": 74.0, "grad_norm": 1.09029221534729, "learning_rate": 1.3000000000000001e-05, "loss": 0.0748, "step": 7844 }, { "epoch": 74.0, "eval_LOCATION_f1": 0.7792207792207793, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7228915662650602, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.738255033557047, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6547619047619048, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8945686900958466, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8588957055214724, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7500000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6923076923076923, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1335461437702179, "eval_overall_accuracy": 0.9602327837051406, "eval_overall_f1": 0.8203753351206434, "eval_overall_precision": 0.7669172932330827, "eval_overall_recall": 0.8818443804034583, "eval_runtime": 0.3744, "eval_samples_per_second": 499.424, "eval_steps_per_second": 8.012, "step": 7844 }, { "epoch": 75.0, "grad_norm": 0.6912786960601807, "learning_rate": 1.25e-05, "loss": 0.0749, "step": 7950 }, { "epoch": 75.0, "eval_LOCATION_f1": 0.7922077922077921, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7349397590361446, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.7346938775510204, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6585365853658537, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.8917197452229298, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8536585365853658, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7500000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6923076923076923, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1301984041929245, "eval_overall_accuracy": 0.9604752667313288, "eval_overall_f1": 0.8214765100671141, "eval_overall_precision": 0.7688442211055276, "eval_overall_recall": 0.8818443804034583, "eval_runtime": 0.3758, "eval_samples_per_second": 497.572, "eval_steps_per_second": 7.982, "step": 7950 }, { "epoch": 76.0, "grad_norm": 1.183219075202942, "learning_rate": 1.2e-05, "loss": 0.0766, "step": 8056 }, { "epoch": 76.0, "eval_LOCATION_f1": 0.7792207792207793, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7228915662650602, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7333333333333334, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6470588235294118, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8952380952380952, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8545454545454545, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7500000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6923076923076923, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13669635355472565, "eval_overall_accuracy": 0.9602327837051406, "eval_overall_f1": 0.8197596795727637, "eval_overall_precision": 0.763681592039801, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3759, "eval_samples_per_second": 497.525, "eval_steps_per_second": 7.982, "step": 8056 }, { "epoch": 77.0, "grad_norm": 3.904420852661133, "learning_rate": 1.1500000000000002e-05, "loss": 0.0766, "step": 8162 }, { "epoch": 77.0, "eval_LOCATION_f1": 0.7792207792207793, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7228915662650602, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7199999999999999, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6352941176470588, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.9096774193548387, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.88125, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13504381477832794, "eval_overall_accuracy": 0.9602327837051406, "eval_overall_f1": 0.8214765100671141, "eval_overall_precision": 0.7688442211055276, "eval_overall_recall": 0.8818443804034583, "eval_runtime": 0.3776, "eval_samples_per_second": 495.232, "eval_steps_per_second": 7.945, "step": 8162 }, { "epoch": 78.0, "grad_norm": 1.537532925605774, "learning_rate": 1.1000000000000001e-05, "loss": 0.0749, "step": 8268 }, { "epoch": 78.0, "eval_LOCATION_f1": 0.7922077922077921, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7349397590361446, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.7333333333333334, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6470588235294118, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9096774193548387, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.88125, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7500000000000001, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6923076923076923, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13401678204536438, "eval_overall_accuracy": 0.960717749757517, "eval_overall_f1": 0.8279569892473119, "eval_overall_precision": 0.7758186397984886, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.3757, "eval_samples_per_second": 497.733, "eval_steps_per_second": 7.985, "step": 8268 }, { "epoch": 79.0, "grad_norm": 1.1565873622894287, "learning_rate": 1.05e-05, "loss": 0.0746, "step": 8374 }, { "epoch": 79.0, "eval_LOCATION_f1": 0.7792207792207793, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7228915662650602, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.728476821192053, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6395348837209303, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9073482428115016, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8711656441717791, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13822628557682037, "eval_overall_accuracy": 0.9597478176527643, "eval_overall_f1": 0.8213333333333332, "eval_overall_precision": 0.7642679900744417, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.3771, "eval_samples_per_second": 495.838, "eval_steps_per_second": 7.955, "step": 8374 }, { "epoch": 80.0, "grad_norm": 1.1117786169052124, "learning_rate": 1e-05, "loss": 0.0736, "step": 8480 }, { "epoch": 80.0, "eval_LOCATION_f1": 0.7792207792207793, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7228915662650602, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7152317880794703, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.627906976744186, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.9137380191693292, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8773006134969326, "eval_PERSON_recall": 0.9533333333333334, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13752157986164093, "eval_overall_accuracy": 0.9602327837051406, "eval_overall_f1": 0.822429906542056, "eval_overall_precision": 0.7661691542288557, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.3752, "eval_samples_per_second": 498.445, "eval_steps_per_second": 7.996, "step": 8480 }, { "epoch": 81.0, "grad_norm": 0.795144259929657, "learning_rate": 9.5e-06, "loss": 0.0732, "step": 8586 }, { "epoch": 81.0, "eval_LOCATION_f1": 0.7870967741935485, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7261904761904762, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.7417218543046358, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6511627906976745, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9161290322580645, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8875, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13531745970249176, "eval_overall_accuracy": 0.9609602327837051, "eval_overall_f1": 0.8299866131191432, "eval_overall_precision": 0.775, "eval_overall_recall": 0.8933717579250721, "eval_runtime": 0.3737, "eval_samples_per_second": 500.356, "eval_steps_per_second": 8.027, "step": 8586 }, { "epoch": 82.0, "grad_norm": 1.324391484260559, "learning_rate": 9e-06, "loss": 0.0757, "step": 8692 }, { "epoch": 82.0, "eval_LOCATION_f1": 0.751592356687898, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.686046511627907, "eval_LOCATION_recall": 0.8309859154929577, "eval_ORGANIZATION_f1": 0.7058823529411765, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6136363636363636, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.9044585987261147, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8658536585365854, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8135593220338982, "eval_TIME_number": 28, "eval_TIME_precision": 0.7741935483870968, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.14012309908866882, "eval_overall_accuracy": 0.9580504364694471, "eval_overall_f1": 0.808454425363276, "eval_overall_precision": 0.7463414634146341, "eval_overall_recall": 0.8818443804034583, "eval_runtime": 0.3745, "eval_samples_per_second": 499.32, "eval_steps_per_second": 8.01, "step": 8692 }, { "epoch": 83.0, "grad_norm": 2.1923840045928955, "learning_rate": 8.500000000000002e-06, "loss": 0.0743, "step": 8798 }, { "epoch": 83.0, "eval_LOCATION_f1": 0.7692307692307694, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7058823529411765, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7152317880794703, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.627906976744186, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.9044585987261147, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8658536585365854, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13606394827365875, "eval_overall_accuracy": 0.9597478176527643, "eval_overall_f1": 0.8164893617021276, "eval_overall_precision": 0.7580246913580246, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3724, "eval_samples_per_second": 502.135, "eval_steps_per_second": 8.056, "step": 8798 }, { "epoch": 84.0, "grad_norm": 1.4156537055969238, "learning_rate": 8.000000000000001e-06, "loss": 0.0761, "step": 8904 }, { "epoch": 84.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.738255033557047, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6547619047619048, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9161290322580645, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8875, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1343669891357422, "eval_overall_accuracy": 0.9604752667313288, "eval_overall_f1": 0.8268456375838926, "eval_overall_precision": 0.7738693467336684, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.3728, "eval_samples_per_second": 501.63, "eval_steps_per_second": 8.048, "step": 8904 }, { "epoch": 85.0, "grad_norm": 2.533181667327881, "learning_rate": 7.5e-06, "loss": 0.0738, "step": 9010 }, { "epoch": 85.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7516778523489933, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6666666666666666, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9131832797427653, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8819875776397516, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13643313944339752, "eval_overall_accuracy": 0.9609602327837051, "eval_overall_f1": 0.8284182305630027, "eval_overall_precision": 0.7744360902255639, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.3727, "eval_samples_per_second": 501.725, "eval_steps_per_second": 8.049, "step": 9010 }, { "epoch": 86.0, "grad_norm": 1.149907112121582, "learning_rate": 7.000000000000001e-06, "loss": 0.0717, "step": 9116 }, { "epoch": 86.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7516778523489933, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6666666666666666, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9096774193548387, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.88125, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7297297297297297, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.6585365853658537, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13564659655094147, "eval_overall_accuracy": 0.9604752667313288, "eval_overall_f1": 0.8257372654155496, "eval_overall_precision": 0.7719298245614035, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.3756, "eval_samples_per_second": 497.883, "eval_steps_per_second": 7.987, "step": 9116 }, { "epoch": 87.0, "grad_norm": 0.7291505932807922, "learning_rate": 6.5000000000000004e-06, "loss": 0.0716, "step": 9222 }, { "epoch": 87.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7466666666666668, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6588235294117647, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9044585987261147, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8658536585365854, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1373915821313858, "eval_overall_accuracy": 0.9599903006789525, "eval_overall_f1": 0.824, "eval_overall_precision": 0.7667493796526055, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.3757, "eval_samples_per_second": 497.799, "eval_steps_per_second": 7.986, "step": 9222 }, { "epoch": 88.0, "grad_norm": 0.8719342350959778, "learning_rate": 6e-06, "loss": 0.0721, "step": 9328 }, { "epoch": 88.0, "eval_LOCATION_f1": 0.7870967741935485, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7261904761904762, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.7466666666666668, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6588235294117647, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9161290322580645, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8875, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7671232876712328, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.7, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1352623850107193, "eval_overall_accuracy": 0.9609602327837051, "eval_overall_f1": 0.8337801608579088, "eval_overall_precision": 0.7794486215538847, "eval_overall_recall": 0.8962536023054755, "eval_runtime": 0.375, "eval_samples_per_second": 498.638, "eval_steps_per_second": 8.0, "step": 9328 }, { "epoch": 89.0, "grad_norm": 1.7583541870117188, "learning_rate": 5.500000000000001e-06, "loss": 0.0736, "step": 9434 }, { "epoch": 89.0, "eval_LOCATION_f1": 0.7870967741935485, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7261904761904762, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.738255033557047, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6547619047619048, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9161290322580645, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8875, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7671232876712328, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.7, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13373106718063354, "eval_overall_accuracy": 0.9609602327837051, "eval_overall_f1": 0.8322147651006712, "eval_overall_precision": 0.7788944723618091, "eval_overall_recall": 0.8933717579250721, "eval_runtime": 0.3865, "eval_samples_per_second": 483.77, "eval_steps_per_second": 7.761, "step": 9434 }, { "epoch": 90.0, "grad_norm": 0.8356399536132812, "learning_rate": 5e-06, "loss": 0.0723, "step": 9540 }, { "epoch": 90.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7516778523489933, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6666666666666666, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9073482428115016, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8711656441717791, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7671232876712328, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.7, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13568884134292603, "eval_overall_accuracy": 0.9612027158098934, "eval_overall_f1": 0.8288770053475935, "eval_overall_precision": 0.773067331670823, "eval_overall_recall": 0.8933717579250721, "eval_runtime": 0.3761, "eval_samples_per_second": 497.201, "eval_steps_per_second": 7.976, "step": 9540 }, { "epoch": 91.0, "grad_norm": 0.7594178915023804, "learning_rate": 4.5e-06, "loss": 0.073, "step": 9646 }, { "epoch": 91.0, "eval_LOCATION_f1": 0.7870967741935485, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7261904761904762, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.738255033557047, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6547619047619048, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9102564102564102, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8765432098765432, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7671232876712328, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.7, "eval_QUANTITY_recall": 0.8484848484848485, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1347140520811081, "eval_overall_accuracy": 0.9609602327837051, "eval_overall_f1": 0.8299866131191432, "eval_overall_precision": 0.775, "eval_overall_recall": 0.8933717579250721, "eval_runtime": 0.3754, "eval_samples_per_second": 498.127, "eval_steps_per_second": 7.991, "step": 9646 }, { "epoch": 92.0, "grad_norm": 2.9121735095977783, "learning_rate": 4.000000000000001e-06, "loss": 0.073, "step": 9752 }, { "epoch": 92.0, "eval_LOCATION_f1": 0.7870967741935485, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7261904761904762, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.7162162162162162, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6385542168674698, "eval_ORGANIZATION_recall": 0.8153846153846154, "eval_PERSON_f1": 0.9073482428115016, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8711656441717791, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13472875952720642, "eval_overall_accuracy": 0.9604752667313288, "eval_overall_f1": 0.821954484605087, "eval_overall_precision": 0.7675, "eval_overall_recall": 0.8847262247838616, "eval_runtime": 0.3767, "eval_samples_per_second": 496.351, "eval_steps_per_second": 7.963, "step": 9752 }, { "epoch": 93.0, "grad_norm": 0.6299099326133728, "learning_rate": 3.5000000000000004e-06, "loss": 0.0725, "step": 9858 }, { "epoch": 93.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7333333333333334, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6470588235294118, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9073482428115016, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8711656441717791, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13586550951004028, "eval_overall_accuracy": 0.9602327837051406, "eval_overall_f1": 0.822429906542056, "eval_overall_precision": 0.7661691542288557, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.3747, "eval_samples_per_second": 499.023, "eval_steps_per_second": 8.006, "step": 9858 }, { "epoch": 94.0, "grad_norm": 1.431706190109253, "learning_rate": 3e-06, "loss": 0.0717, "step": 9964 }, { "epoch": 94.0, "eval_LOCATION_f1": 0.7692307692307694, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7058823529411765, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7333333333333334, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6470588235294118, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9044585987261147, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8658536585365854, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1379024237394333, "eval_overall_accuracy": 0.9597478176527643, "eval_overall_f1": 0.8202396804260986, "eval_overall_precision": 0.7623762376237624, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.3743, "eval_samples_per_second": 499.653, "eval_steps_per_second": 8.016, "step": 9964 }, { "epoch": 95.0, "grad_norm": 1.1127707958221436, "learning_rate": 2.5e-06, "loss": 0.0725, "step": 10070 }, { "epoch": 95.0, "eval_LOCATION_f1": 0.7692307692307694, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7058823529411765, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7466666666666668, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6588235294117647, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9073482428115016, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8711656441717791, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1367916464805603, "eval_overall_accuracy": 0.9602327837051406, "eval_overall_f1": 0.824, "eval_overall_precision": 0.7667493796526055, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.3746, "eval_samples_per_second": 499.244, "eval_steps_per_second": 8.009, "step": 10070 }, { "epoch": 96.0, "grad_norm": 1.2149391174316406, "learning_rate": 2.0000000000000003e-06, "loss": 0.0707, "step": 10176 }, { "epoch": 96.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7466666666666668, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6588235294117647, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9102564102564102, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8765432098765432, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1361166536808014, "eval_overall_accuracy": 0.960717749757517, "eval_overall_f1": 0.8262032085561497, "eval_overall_precision": 0.770573566084788, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.3758, "eval_samples_per_second": 497.542, "eval_steps_per_second": 7.982, "step": 10176 }, { "epoch": 97.0, "grad_norm": 1.4393696784973145, "learning_rate": 1.5e-06, "loss": 0.0724, "step": 10282 }, { "epoch": 97.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7333333333333334, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6470588235294118, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9102564102564102, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8765432098765432, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13554507493972778, "eval_overall_accuracy": 0.9604752667313288, "eval_overall_f1": 0.823529411764706, "eval_overall_precision": 0.7680798004987531, "eval_overall_recall": 0.8876080691642652, "eval_runtime": 0.3752, "eval_samples_per_second": 498.429, "eval_steps_per_second": 7.996, "step": 10282 }, { "epoch": 98.0, "grad_norm": 1.7291535139083862, "learning_rate": 1.0000000000000002e-06, "loss": 0.0708, "step": 10388 }, { "epoch": 98.0, "eval_LOCATION_f1": 0.7870967741935485, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7261904761904762, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.7466666666666668, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6588235294117647, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9131832797427653, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8819875776397516, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13509468734264374, "eval_overall_accuracy": 0.9612027158098934, "eval_overall_f1": 0.8299866131191432, "eval_overall_precision": 0.775, "eval_overall_recall": 0.8933717579250721, "eval_runtime": 0.3737, "eval_samples_per_second": 500.414, "eval_steps_per_second": 8.028, "step": 10388 }, { "epoch": 99.0, "grad_norm": 1.2517259120941162, "learning_rate": 5.000000000000001e-07, "loss": 0.0725, "step": 10494 }, { "epoch": 99.0, "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8450704225352113, "eval_ORGANIZATION_f1": 0.7466666666666668, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6588235294117647, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9102564102564102, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8765432098765432, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.1357007920742035, "eval_overall_accuracy": 0.960717749757517, "eval_overall_f1": 0.8262032085561497, "eval_overall_precision": 0.770573566084788, "eval_overall_recall": 0.8904899135446686, "eval_runtime": 0.376, "eval_samples_per_second": 497.368, "eval_steps_per_second": 7.979, "step": 10494 }, { "epoch": 100.0, "grad_norm": 0.5946673154830933, "learning_rate": 0.0, "loss": 0.0719, "step": 10600 }, { "epoch": 100.0, "eval_LOCATION_f1": 0.7870967741935485, "eval_LOCATION_number": 71, "eval_LOCATION_precision": 0.7261904761904762, "eval_LOCATION_recall": 0.8591549295774648, "eval_ORGANIZATION_f1": 0.7466666666666668, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6588235294117647, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9102564102564102, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8765432098765432, "eval_PERSON_recall": 0.9466666666666667, "eval_QUANTITY_f1": 0.7397260273972603, "eval_QUANTITY_number": 33, "eval_QUANTITY_precision": 0.675, "eval_QUANTITY_recall": 0.8181818181818182, "eval_TIME_f1": 0.8275862068965518, "eval_TIME_number": 28, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8571428571428571, "eval_loss": 0.13545459508895874, "eval_overall_accuracy": 0.9609602327837051, "eval_overall_f1": 0.8288770053475935, "eval_overall_precision": 0.773067331670823, "eval_overall_recall": 0.8933717579250721, "eval_runtime": 0.3737, "eval_samples_per_second": 500.414, "eval_steps_per_second": 8.028, "step": 10600 }, { "epoch": 100.0, "step": 10600, "total_flos": 4623411255592272.0, "train_loss": 0.1278715612303536, "train_runtime": 581.8892, "train_samples_per_second": 290.09, "train_steps_per_second": 18.217 } ], "logging_steps": 500, "max_steps": 10600, "num_input_tokens_seen": 0, "num_train_epochs": 100, "save_steps": 500, "total_flos": 4623411255592272.0, "train_batch_size": 16, "trial_name": null, "trial_params": null }