{ "best_metric": null, "best_model_checkpoint": null, "epoch": 100.0, "eval_steps": 500, "global_step": 10600, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 1.0, "grad_norm": 1.9069651365280151, "learning_rate": 4.9500000000000004e-05, "loss": 0.9989, "step": 106 }, { "epoch": 1.0, "eval_LOCATION_f1": 0.0, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.0, "eval_LOCATION_recall": 0.0, "eval_ORGANIZATION_f1": 0.0, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.0, "eval_ORGANIZATION_recall": 0.0, "eval_PERSON_f1": 0.012987012987012986, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.09090909090909091, "eval_PERSON_recall": 0.006993006993006993, "eval_QUANTITY_f1": 0.0, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.0, "eval_QUANTITY_recall": 0.0, "eval_TIME_f1": 0.0, "eval_TIME_number": 26, "eval_TIME_precision": 0.0, "eval_TIME_recall": 0.0, "eval_loss": 0.5582299828529358, "eval_overall_accuracy": 0.8468851638729047, "eval_overall_f1": 0.00554016620498615, "eval_overall_precision": 0.045454545454545456, "eval_overall_recall": 0.0029498525073746312, "eval_runtime": 0.3749, "eval_samples_per_second": 498.785, "eval_steps_per_second": 8.002, "step": 106 }, { "epoch": 2.0, "grad_norm": 1.1202826499938965, "learning_rate": 4.9e-05, "loss": 0.4921, "step": 212 }, { "epoch": 2.0, "eval_LOCATION_f1": 0.20618556701030927, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.4, "eval_LOCATION_recall": 0.1388888888888889, "eval_ORGANIZATION_f1": 0.03389830508474576, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.046511627906976744, "eval_ORGANIZATION_recall": 0.02666666666666667, "eval_PERSON_f1": 0.5795454545454545, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.4880382775119617, "eval_PERSON_recall": 0.7132867132867133, "eval_QUANTITY_f1": 0.11764705882352941, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.10714285714285714, "eval_QUANTITY_recall": 0.13043478260869565, "eval_TIME_f1": 0.6, "eval_TIME_number": 26, "eval_TIME_precision": 0.625, "eval_TIME_recall": 0.5769230769230769, "eval_loss": 0.328204870223999, "eval_overall_accuracy": 0.9014260695521641, "eval_overall_f1": 0.39520958083832336, "eval_overall_precision": 0.4012158054711246, "eval_overall_recall": 0.3893805309734513, "eval_runtime": 0.3804, "eval_samples_per_second": 491.588, "eval_steps_per_second": 7.886, "step": 212 }, { "epoch": 3.0, "grad_norm": 1.1920843124389648, "learning_rate": 4.85e-05, "loss": 0.3179, "step": 318 }, { "epoch": 3.0, "eval_LOCATION_f1": 0.6705882352941176, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.5816326530612245, "eval_LOCATION_recall": 0.7916666666666666, "eval_ORGANIZATION_f1": 0.48314606741573035, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.4174757281553398, "eval_ORGANIZATION_recall": 0.5733333333333334, "eval_PERSON_f1": 0.8295819935691319, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.7678571428571429, "eval_PERSON_recall": 0.9020979020979021, "eval_QUANTITY_f1": 0.4545454545454546, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.3488372093023256, "eval_QUANTITY_recall": 0.6521739130434783, "eval_TIME_f1": 0.7368421052631579, "eval_TIME_number": 26, "eval_TIME_precision": 0.6774193548387096, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.23815979063510895, "eval_overall_accuracy": 0.92694520890668, "eval_overall_f1": 0.6777493606138107, "eval_overall_precision": 0.5981941309255079, "eval_overall_recall": 0.7817109144542773, "eval_runtime": 0.3833, "eval_samples_per_second": 487.894, "eval_steps_per_second": 7.827, "step": 318 }, { "epoch": 4.0, "grad_norm": 1.0607746839523315, "learning_rate": 4.8e-05, "loss": 0.2251, "step": 424 }, { "epoch": 4.0, "eval_LOCATION_f1": 0.7250000000000001, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.6590909090909091, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.6091954022988506, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.5353535353535354, "eval_ORGANIZATION_recall": 0.7066666666666667, "eval_PERSON_f1": 0.8417721518987342, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.7687861271676301, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.47058823529411764, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.35555555555555557, "eval_QUANTITY_recall": 0.6956521739130435, "eval_TIME_f1": 0.6666666666666666, "eval_TIME_number": 26, "eval_TIME_precision": 0.5675675675675675, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.21234844624996185, "eval_overall_accuracy": 0.9321991493620215, "eval_overall_f1": 0.7195902688860435, "eval_overall_precision": 0.6357466063348416, "eval_overall_recall": 0.8289085545722714, "eval_runtime": 0.385, "eval_samples_per_second": 485.682, "eval_steps_per_second": 7.792, "step": 424 }, { "epoch": 5.0, "grad_norm": 1.1459556818008423, "learning_rate": 4.75e-05, "loss": 0.1854, "step": 530 }, { "epoch": 5.0, "eval_LOCATION_f1": 0.7333333333333334, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7051282051282052, "eval_LOCATION_recall": 0.7638888888888888, "eval_ORGANIZATION_f1": 0.6589595375722543, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.5816326530612245, "eval_ORGANIZATION_recall": 0.76, "eval_PERSON_f1": 0.8571428571428572, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8164556962025317, "eval_PERSON_recall": 0.9020979020979021, "eval_QUANTITY_f1": 0.5862068965517241, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.4857142857142857, "eval_QUANTITY_recall": 0.7391304347826086, "eval_TIME_f1": 0.8679245283018868, "eval_TIME_number": 26, "eval_TIME_precision": 0.8518518518518519, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.1583337038755417, "eval_overall_accuracy": 0.9484613460095072, "eval_overall_f1": 0.7646258503401361, "eval_overall_precision": 0.7095959595959596, "eval_overall_recall": 0.8289085545722714, "eval_runtime": 0.382, "eval_samples_per_second": 489.538, "eval_steps_per_second": 7.854, "step": 530 }, { "epoch": 6.0, "grad_norm": 1.156198263168335, "learning_rate": 4.7e-05, "loss": 0.1674, "step": 636 }, { "epoch": 6.0, "eval_LOCATION_f1": 0.7065868263473054, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.6210526315789474, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.6703296703296704, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.5700934579439252, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.8947368421052632, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.84472049689441, "eval_PERSON_recall": 0.951048951048951, "eval_QUANTITY_f1": 0.5964912280701754, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5, "eval_QUANTITY_recall": 0.7391304347826086, "eval_TIME_f1": 0.8571428571428571, "eval_TIME_number": 26, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.9230769230769231, "eval_loss": 0.15457449853420258, "eval_overall_accuracy": 0.9492119089316988, "eval_overall_f1": 0.7754569190600523, "eval_overall_precision": 0.6955503512880562, "eval_overall_recall": 0.8761061946902655, "eval_runtime": 0.3837, "eval_samples_per_second": 487.343, "eval_steps_per_second": 7.818, "step": 636 }, { "epoch": 7.0, "grad_norm": 0.5854206681251526, "learning_rate": 4.6500000000000005e-05, "loss": 0.1515, "step": 742 }, { "epoch": 7.0, "eval_LOCATION_f1": 0.751592356687898, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.6941176470588235, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.6783625730994152, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6041666666666666, "eval_ORGANIZATION_recall": 0.7733333333333333, "eval_PERSON_f1": 0.8888888888888888, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8343558282208589, "eval_PERSON_recall": 0.951048951048951, "eval_QUANTITY_f1": 0.5806451612903226, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.46153846153846156, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.7931034482758621, "eval_TIME_number": 26, "eval_TIME_precision": 0.71875, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.15316547453403473, "eval_overall_accuracy": 0.9512134100575431, "eval_overall_f1": 0.7798408488063661, "eval_overall_precision": 0.708433734939759, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.383, "eval_samples_per_second": 488.261, "eval_steps_per_second": 7.833, "step": 742 }, { "epoch": 8.0, "grad_norm": 0.6810721158981323, "learning_rate": 4.600000000000001e-05, "loss": 0.1437, "step": 848 }, { "epoch": 8.0, "eval_LOCATION_f1": 0.7435897435897436, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.6904761904761905, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.6823529411764706, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6105263157894737, "eval_ORGANIZATION_recall": 0.7733333333333333, "eval_PERSON_f1": 0.8947368421052632, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.84472049689441, "eval_PERSON_recall": 0.951048951048951, "eval_QUANTITY_f1": 0.6666666666666667, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5588235294117647, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7457627118644068, "eval_TIME_number": 26, "eval_TIME_precision": 0.6666666666666666, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.14404982328414917, "eval_overall_accuracy": 0.9519639729797348, "eval_overall_f1": 0.7855227882037534, "eval_overall_precision": 0.7199017199017199, "eval_overall_recall": 0.8643067846607669, "eval_runtime": 0.3839, "eval_samples_per_second": 487.113, "eval_steps_per_second": 7.815, "step": 848 }, { "epoch": 9.0, "grad_norm": 0.8699572086334229, "learning_rate": 4.55e-05, "loss": 0.1356, "step": 954 }, { "epoch": 9.0, "eval_LOCATION_f1": 0.7450980392156864, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7037037037037037, "eval_LOCATION_recall": 0.7916666666666666, "eval_ORGANIZATION_f1": 0.7052023121387283, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6224489795918368, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.9013157894736843, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8509316770186336, "eval_PERSON_recall": 0.958041958041958, "eval_QUANTITY_f1": 0.7307692307692308, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6551724137931034, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.75, "eval_TIME_number": 26, "eval_TIME_precision": 0.7, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.14066973328590393, "eval_overall_accuracy": 0.9537152864648486, "eval_overall_f1": 0.7994579945799458, "eval_overall_precision": 0.7393483709273183, "eval_overall_recall": 0.8702064896755162, "eval_runtime": 0.379, "eval_samples_per_second": 493.436, "eval_steps_per_second": 7.916, "step": 954 }, { "epoch": 10.0, "grad_norm": 2.338326930999756, "learning_rate": 4.5e-05, "loss": 0.1257, "step": 1060 }, { "epoch": 10.0, "eval_LOCATION_f1": 0.7733333333333334, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7435897435897436, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.732919254658385, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.686046511627907, "eval_ORGANIZATION_recall": 0.7866666666666666, "eval_PERSON_f1": 0.8903654485049833, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8481012658227848, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.5964912280701754, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5, "eval_QUANTITY_recall": 0.7391304347826086, "eval_TIME_f1": 0.7586206896551724, "eval_TIME_number": 26, "eval_TIME_precision": 0.6875, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.13949567079544067, "eval_overall_accuracy": 0.9537152864648486, "eval_overall_f1": 0.797799174690509, "eval_overall_precision": 0.7474226804123711, "eval_overall_recall": 0.855457227138643, "eval_runtime": 0.379, "eval_samples_per_second": 493.451, "eval_steps_per_second": 7.916, "step": 1060 }, { "epoch": 11.0, "grad_norm": 0.916523277759552, "learning_rate": 4.4500000000000004e-05, "loss": 0.1201, "step": 1166 }, { "epoch": 11.0, "eval_LOCATION_f1": 0.758169934640523, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7160493827160493, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.7682926829268292, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7078651685393258, "eval_ORGANIZATION_recall": 0.84, "eval_PERSON_f1": 0.8851351351351351, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8562091503267973, "eval_PERSON_recall": 0.916083916083916, "eval_QUANTITY_f1": 0.6551724137931034, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5428571428571428, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8070175438596492, "eval_TIME_number": 26, "eval_TIME_precision": 0.7419354838709677, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.13555747270584106, "eval_overall_accuracy": 0.9549662246685013, "eval_overall_f1": 0.8076923076923076, "eval_overall_precision": 0.7557840616966581, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.3787, "eval_samples_per_second": 493.763, "eval_steps_per_second": 7.921, "step": 1166 }, { "epoch": 12.0, "grad_norm": 1.767835021018982, "learning_rate": 4.4000000000000006e-05, "loss": 0.1165, "step": 1272 }, { "epoch": 12.0, "eval_LOCATION_f1": 0.7651006711409396, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7402597402597403, "eval_LOCATION_recall": 0.7916666666666666, "eval_ORGANIZATION_f1": 0.7826086956521738, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7325581395348837, "eval_ORGANIZATION_recall": 0.84, "eval_PERSON_f1": 0.8903654485049833, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8481012658227848, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.6181818181818182, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.53125, "eval_QUANTITY_recall": 0.7391304347826086, "eval_TIME_f1": 0.9230769230769231, "eval_TIME_number": 26, "eval_TIME_precision": 0.9230769230769231, "eval_TIME_recall": 0.9230769230769231, "eval_loss": 0.13143397867679596, "eval_overall_accuracy": 0.9589692269201902, "eval_overall_f1": 0.8217270194986072, "eval_overall_precision": 0.7783641160949868, "eval_overall_recall": 0.8702064896755162, "eval_runtime": 0.3789, "eval_samples_per_second": 493.532, "eval_steps_per_second": 7.918, "step": 1272 }, { "epoch": 13.0, "grad_norm": 0.5636658072471619, "learning_rate": 4.35e-05, "loss": 0.1099, "step": 1378 }, { "epoch": 13.0, "eval_LOCATION_f1": 0.7712418300653595, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7283950617283951, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.7607361963190183, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7045454545454546, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.8963210702341137, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8589743589743589, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7169811320754716, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6333333333333333, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8571428571428571, "eval_TIME_number": 26, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.9230769230769231, "eval_loss": 0.13585500419139862, "eval_overall_accuracy": 0.9589692269201902, "eval_overall_f1": 0.8232044198895028, "eval_overall_precision": 0.7740259740259741, "eval_overall_recall": 0.8790560471976401, "eval_runtime": 0.3821, "eval_samples_per_second": 489.344, "eval_steps_per_second": 7.85, "step": 1378 }, { "epoch": 14.0, "grad_norm": 0.9262148141860962, "learning_rate": 4.3e-05, "loss": 0.1083, "step": 1484 }, { "epoch": 14.0, "eval_LOCATION_f1": 0.7375886524822696, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7536231884057971, "eval_LOCATION_recall": 0.7222222222222222, "eval_ORGANIZATION_f1": 0.7784431137724551, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7065217391304348, "eval_ORGANIZATION_recall": 0.8666666666666667, "eval_PERSON_f1": 0.8903654485049833, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8481012658227848, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.6545454545454547, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5625, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.7368421052631579, "eval_TIME_number": 26, "eval_TIME_precision": 0.6774193548387096, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.13395123183727264, "eval_overall_accuracy": 0.9567175381536153, "eval_overall_f1": 0.8044382801664355, "eval_overall_precision": 0.7591623036649214, "eval_overall_recall": 0.855457227138643, "eval_runtime": 0.3776, "eval_samples_per_second": 495.296, "eval_steps_per_second": 7.946, "step": 1484 }, { "epoch": 15.0, "grad_norm": 0.5211468935012817, "learning_rate": 4.25e-05, "loss": 0.1016, "step": 1590 }, { "epoch": 15.0, "eval_LOCATION_f1": 0.7733333333333334, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7435897435897436, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.7619047619047621, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6881720430107527, "eval_ORGANIZATION_recall": 0.8533333333333334, "eval_PERSON_f1": 0.8963210702341137, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8589743589743589, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8571428571428571, "eval_TIME_number": 26, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.9230769230769231, "eval_loss": 0.132180854678154, "eval_overall_accuracy": 0.9594696022016512, "eval_overall_f1": 0.8271092669432918, "eval_overall_precision": 0.7786458333333334, "eval_overall_recall": 0.8820058997050148, "eval_runtime": 0.3807, "eval_samples_per_second": 491.223, "eval_steps_per_second": 7.881, "step": 1590 }, { "epoch": 16.0, "grad_norm": 0.5465927720069885, "learning_rate": 4.2e-05, "loss": 0.0981, "step": 1696 }, { "epoch": 16.0, "eval_LOCATION_f1": 0.781456953642384, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7468354430379747, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.7457627118644068, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6470588235294118, "eval_ORGANIZATION_recall": 0.88, "eval_PERSON_f1": 0.903654485049834, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8607594936708861, "eval_PERSON_recall": 0.951048951048951, "eval_QUANTITY_f1": 0.7058823529411765, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6428571428571429, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8214285714285715, "eval_TIME_number": 26, "eval_TIME_precision": 0.7666666666666667, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.14411497116088867, "eval_overall_accuracy": 0.9567175381536153, "eval_overall_f1": 0.8206521739130435, "eval_overall_precision": 0.760705289672544, "eval_overall_recall": 0.8908554572271387, "eval_runtime": 0.3812, "eval_samples_per_second": 490.506, "eval_steps_per_second": 7.869, "step": 1696 }, { "epoch": 17.0, "grad_norm": 0.35876792669296265, "learning_rate": 4.15e-05, "loss": 0.0971, "step": 1802 }, { "epoch": 17.0, "eval_LOCATION_f1": 0.75, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7125, "eval_LOCATION_recall": 0.7916666666666666, "eval_ORGANIZATION_f1": 0.7804878048780487, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7191011235955056, "eval_ORGANIZATION_recall": 0.8533333333333334, "eval_PERSON_f1": 0.8963210702341137, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8589743589743589, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.6666666666666667, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5806451612903226, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.7931034482758621, "eval_TIME_number": 26, "eval_TIME_precision": 0.71875, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.14179494976997375, "eval_overall_accuracy": 0.9572179134350763, "eval_overall_f1": 0.8143053645116919, "eval_overall_precision": 0.7628865979381443, "eval_overall_recall": 0.8731563421828908, "eval_runtime": 0.3784, "eval_samples_per_second": 494.134, "eval_steps_per_second": 7.927, "step": 1802 }, { "epoch": 18.0, "grad_norm": 0.45972150564193726, "learning_rate": 4.1e-05, "loss": 0.0915, "step": 1908 }, { "epoch": 18.0, "eval_LOCATION_f1": 0.7516778523489932, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7272727272727273, "eval_LOCATION_recall": 0.7777777777777778, "eval_ORGANIZATION_f1": 0.751592356687898, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7195121951219512, "eval_ORGANIZATION_recall": 0.7866666666666666, "eval_PERSON_f1": 0.8866666666666668, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8471337579617835, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7037037037037037, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6129032258064516, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8, "eval_TIME_number": 26, "eval_TIME_precision": 0.7586206896551724, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.1389806866645813, "eval_overall_accuracy": 0.9584688516387291, "eval_overall_f1": 0.8083916083916084, "eval_overall_precision": 0.7686170212765957, "eval_overall_recall": 0.8525073746312685, "eval_runtime": 0.3803, "eval_samples_per_second": 491.697, "eval_steps_per_second": 7.888, "step": 1908 }, { "epoch": 19.0, "grad_norm": 0.869144856929779, "learning_rate": 4.05e-05, "loss": 0.0887, "step": 2014 }, { "epoch": 19.0, "eval_LOCATION_f1": 0.778523489932886, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7532467532467533, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.783625730994152, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6979166666666666, "eval_ORGANIZATION_recall": 0.8933333333333333, "eval_PERSON_f1": 0.8829431438127091, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8461538461538461, "eval_PERSON_recall": 0.9230769230769231, "eval_QUANTITY_f1": 0.6428571428571429, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5454545454545454, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8214285714285715, "eval_TIME_number": 26, "eval_TIME_precision": 0.7666666666666667, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.14166148006916046, "eval_overall_accuracy": 0.9574681010758068, "eval_overall_f1": 0.8153214774281805, "eval_overall_precision": 0.7602040816326531, "eval_overall_recall": 0.8790560471976401, "eval_runtime": 0.3786, "eval_samples_per_second": 493.99, "eval_steps_per_second": 7.925, "step": 2014 }, { "epoch": 20.0, "grad_norm": 0.5048514604568481, "learning_rate": 4e-05, "loss": 0.085, "step": 2120 }, { "epoch": 20.0, "eval_LOCATION_f1": 0.7973856209150327, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7530864197530864, "eval_LOCATION_recall": 0.8472222222222222, "eval_ORGANIZATION_f1": 0.7514450867052023, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6632653061224489, "eval_ORGANIZATION_recall": 0.8666666666666667, "eval_PERSON_f1": 0.9006622516556291, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8553459119496856, "eval_PERSON_recall": 0.951048951048951, "eval_QUANTITY_f1": 0.6923076923076923, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6206896551724138, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8070175438596492, "eval_TIME_number": 26, "eval_TIME_precision": 0.7419354838709677, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.1567302793264389, "eval_overall_accuracy": 0.955716787590693, "eval_overall_f1": 0.8222523744911804, "eval_overall_precision": 0.7613065326633166, "eval_overall_recall": 0.8938053097345132, "eval_runtime": 0.3789, "eval_samples_per_second": 493.58, "eval_steps_per_second": 7.918, "step": 2120 }, { "epoch": 21.0, "grad_norm": 1.0619794130325317, "learning_rate": 3.9500000000000005e-05, "loss": 0.0818, "step": 2226 }, { "epoch": 21.0, "eval_LOCATION_f1": 0.7862068965517242, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7808219178082192, "eval_LOCATION_recall": 0.7916666666666666, "eval_ORGANIZATION_f1": 0.7380952380952381, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6666666666666666, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.9, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8598726114649682, "eval_PERSON_recall": 0.9440559440559441, "eval_QUANTITY_f1": 0.6666666666666667, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5806451612903226, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8518518518518519, "eval_TIME_number": 26, "eval_TIME_precision": 0.8214285714285714, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.14805683493614197, "eval_overall_accuracy": 0.9582186639979985, "eval_overall_f1": 0.8183079056865463, "eval_overall_precision": 0.7722513089005235, "eval_overall_recall": 0.8702064896755162, "eval_runtime": 0.3803, "eval_samples_per_second": 491.687, "eval_steps_per_second": 7.888, "step": 2226 }, { "epoch": 22.0, "grad_norm": 0.8177421689033508, "learning_rate": 3.9000000000000006e-05, "loss": 0.081, "step": 2332 }, { "epoch": 22.0, "eval_LOCATION_f1": 0.8082191780821918, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7972972972972973, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.7311827956989246, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6126126126126126, "eval_ORGANIZATION_recall": 0.9066666666666666, "eval_PERSON_f1": 0.8963210702341137, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8589743589743589, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.6666666666666667, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5806451612903226, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8, "eval_TIME_number": 26, "eval_TIME_precision": 0.7586206896551724, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.15910604596138, "eval_overall_accuracy": 0.9542156617463098, "eval_overall_f1": 0.8135135135135134, "eval_overall_precision": 0.7506234413965087, "eval_overall_recall": 0.887905604719764, "eval_runtime": 0.3796, "eval_samples_per_second": 492.677, "eval_steps_per_second": 7.904, "step": 2332 }, { "epoch": 23.0, "grad_norm": 0.4929425120353699, "learning_rate": 3.85e-05, "loss": 0.0781, "step": 2438 }, { "epoch": 23.0, "eval_LOCATION_f1": 0.7746478873239437, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7857142857142857, "eval_LOCATION_recall": 0.7638888888888888, "eval_ORGANIZATION_f1": 0.7439024390243902, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6853932584269663, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.903010033444816, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8653846153846154, "eval_PERSON_recall": 0.9440559440559441, "eval_QUANTITY_f1": 0.6666666666666667, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5806451612903226, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8363636363636363, "eval_TIME_number": 26, "eval_TIME_precision": 0.7931034482758621, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.14332479238510132, "eval_overall_accuracy": 0.9584688516387291, "eval_overall_f1": 0.8179271708683472, "eval_overall_precision": 0.7786666666666666, "eval_overall_recall": 0.8613569321533924, "eval_runtime": 0.3798, "eval_samples_per_second": 492.33, "eval_steps_per_second": 7.898, "step": 2438 }, { "epoch": 24.0, "grad_norm": 0.7625645995140076, "learning_rate": 3.8e-05, "loss": 0.0762, "step": 2544 }, { "epoch": 24.0, "eval_LOCATION_f1": 0.778523489932886, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7532467532467533, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.782051282051282, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7530864197530864, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.8933333333333333, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8535031847133758, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.6792452830188679, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8148148148148148, "eval_TIME_number": 26, "eval_TIME_precision": 0.7857142857142857, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.1436605453491211, "eval_overall_accuracy": 0.9609707280460346, "eval_overall_f1": 0.8230337078651685, "eval_overall_precision": 0.7855227882037533, "eval_overall_recall": 0.8643067846607669, "eval_runtime": 0.3785, "eval_samples_per_second": 494.087, "eval_steps_per_second": 7.927, "step": 2544 }, { "epoch": 25.0, "grad_norm": 0.27462002635002136, "learning_rate": 3.7500000000000003e-05, "loss": 0.0758, "step": 2650 }, { "epoch": 25.0, "eval_LOCATION_f1": 0.761904761904762, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7466666666666667, "eval_LOCATION_recall": 0.7777777777777778, "eval_ORGANIZATION_f1": 0.7555555555555555, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6476190476190476, "eval_ORGANIZATION_recall": 0.9066666666666666, "eval_PERSON_f1": 0.8963210702341137, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8589743589743589, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.6428571428571429, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5454545454545454, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8214285714285715, "eval_TIME_number": 26, "eval_TIME_precision": 0.7666666666666667, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.1509406715631485, "eval_overall_accuracy": 0.9574681010758068, "eval_overall_f1": 0.8102981029810299, "eval_overall_precision": 0.7493734335839599, "eval_overall_recall": 0.8820058997050148, "eval_runtime": 0.3782, "eval_samples_per_second": 494.445, "eval_steps_per_second": 7.932, "step": 2650 }, { "epoch": 26.0, "grad_norm": 0.7901753187179565, "learning_rate": 3.7e-05, "loss": 0.0739, "step": 2756 }, { "epoch": 26.0, "eval_LOCATION_f1": 0.7919463087248323, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7662337662337663, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.782051282051282, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7530864197530864, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.888888888888889, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8571428571428571, "eval_PERSON_recall": 0.9230769230769231, "eval_QUANTITY_f1": 0.6428571428571429, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5454545454545454, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.830188679245283, "eval_TIME_number": 26, "eval_TIME_precision": 0.8148148148148148, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.14178428053855896, "eval_overall_accuracy": 0.9604703527645734, "eval_overall_f1": 0.8213783403656821, "eval_overall_precision": 0.7849462365591398, "eval_overall_recall": 0.8613569321533924, "eval_runtime": 0.3808, "eval_samples_per_second": 491.068, "eval_steps_per_second": 7.878, "step": 2756 }, { "epoch": 27.0, "grad_norm": 0.5350838303565979, "learning_rate": 3.65e-05, "loss": 0.0712, "step": 2862 }, { "epoch": 27.0, "eval_LOCATION_f1": 0.7651006711409396, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7402597402597403, "eval_LOCATION_recall": 0.7916666666666666, "eval_ORGANIZATION_f1": 0.782051282051282, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7530864197530864, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.8837209302325583, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8417721518987342, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.6538461538461539, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5862068965517241, "eval_QUANTITY_recall": 0.7391304347826086, "eval_TIME_f1": 0.7931034482758621, "eval_TIME_number": 26, "eval_TIME_precision": 0.71875, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.14676761627197266, "eval_overall_accuracy": 0.9582186639979985, "eval_overall_f1": 0.8128491620111733, "eval_overall_precision": 0.7718832891246684, "eval_overall_recall": 0.8584070796460177, "eval_runtime": 0.3793, "eval_samples_per_second": 493.026, "eval_steps_per_second": 7.91, "step": 2862 }, { "epoch": 28.0, "grad_norm": 1.0198100805282593, "learning_rate": 3.6e-05, "loss": 0.0698, "step": 2968 }, { "epoch": 28.0, "eval_LOCATION_f1": 0.7692307692307692, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7746478873239436, "eval_LOCATION_recall": 0.7638888888888888, "eval_ORGANIZATION_f1": 0.7741935483870969, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.75, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8926174496644295, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8580645161290322, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8214285714285715, "eval_TIME_number": 26, "eval_TIME_precision": 0.7666666666666667, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.1400754153728485, "eval_overall_accuracy": 0.9609707280460346, "eval_overall_f1": 0.825035561877667, "eval_overall_precision": 0.7967032967032966, "eval_overall_recall": 0.855457227138643, "eval_runtime": 0.3797, "eval_samples_per_second": 492.554, "eval_steps_per_second": 7.902, "step": 2968 }, { "epoch": 29.0, "grad_norm": 0.5019887685775757, "learning_rate": 3.55e-05, "loss": 0.0676, "step": 3074 }, { "epoch": 29.0, "eval_LOCATION_f1": 0.7945205479452055, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7837837837837838, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.7529411764705883, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6736842105263158, "eval_ORGANIZATION_recall": 0.8533333333333334, "eval_PERSON_f1": 0.8986486486486487, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.869281045751634, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.6792452830188679, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.15197622776031494, "eval_overall_accuracy": 0.9564673505128847, "eval_overall_f1": 0.817174515235457, "eval_overall_precision": 0.7702349869451697, "eval_overall_recall": 0.8702064896755162, "eval_runtime": 0.3798, "eval_samples_per_second": 492.417, "eval_steps_per_second": 7.9, "step": 3074 }, { "epoch": 30.0, "grad_norm": 0.6126108169555664, "learning_rate": 3.5e-05, "loss": 0.0654, "step": 3180 }, { "epoch": 30.0, "eval_LOCATION_f1": 0.7919463087248323, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7662337662337663, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.779874213836478, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7380952380952381, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.888888888888889, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8571428571428571, "eval_PERSON_recall": 0.9230769230769231, "eval_QUANTITY_f1": 0.7058823529411765, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6428571428571429, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8518518518518519, "eval_TIME_number": 26, "eval_TIME_precision": 0.8214285714285714, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.1480516791343689, "eval_overall_accuracy": 0.9617212909682261, "eval_overall_f1": 0.828169014084507, "eval_overall_precision": 0.7924528301886793, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.3789, "eval_samples_per_second": 493.589, "eval_steps_per_second": 7.919, "step": 3180 }, { "epoch": 31.0, "grad_norm": 0.4918847382068634, "learning_rate": 3.45e-05, "loss": 0.0652, "step": 3286 }, { "epoch": 31.0, "eval_LOCATION_f1": 0.7916666666666666, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7916666666666666, "eval_LOCATION_recall": 0.7916666666666666, "eval_ORGANIZATION_f1": 0.7999999999999999, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7529411764705882, "eval_ORGANIZATION_recall": 0.8533333333333334, "eval_PERSON_f1": 0.9054054054054055, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8758169934640523, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8070175438596492, "eval_TIME_number": 26, "eval_TIME_precision": 0.7419354838709677, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.14855852723121643, "eval_overall_accuracy": 0.9617212909682261, "eval_overall_f1": 0.8389830508474576, "eval_overall_precision": 0.8048780487804879, "eval_overall_recall": 0.8761061946902655, "eval_runtime": 0.3827, "eval_samples_per_second": 488.589, "eval_steps_per_second": 7.838, "step": 3286 }, { "epoch": 32.0, "grad_norm": 1.0412123203277588, "learning_rate": 3.4000000000000007e-05, "loss": 0.0622, "step": 3392 }, { "epoch": 32.0, "eval_LOCATION_f1": 0.7702702702702704, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.75, "eval_LOCATION_recall": 0.7916666666666666, "eval_ORGANIZATION_f1": 0.7701863354037267, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7209302325581395, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.8993288590604027, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.864516129032258, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.6923076923076923, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6206896551724138, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8474576271186441, "eval_TIME_number": 26, "eval_TIME_precision": 0.7575757575757576, "eval_TIME_recall": 0.9615384615384616, "eval_loss": 0.15195608139038086, "eval_overall_accuracy": 0.9599699774831123, "eval_overall_f1": 0.8245125348189415, "eval_overall_precision": 0.7810026385224275, "eval_overall_recall": 0.8731563421828908, "eval_runtime": 0.3788, "eval_samples_per_second": 493.704, "eval_steps_per_second": 7.92, "step": 3392 }, { "epoch": 33.0, "grad_norm": 1.087945580482483, "learning_rate": 3.35e-05, "loss": 0.06, "step": 3498 }, { "epoch": 33.0, "eval_LOCATION_f1": 0.8055555555555556, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8055555555555556, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.744186046511628, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6597938144329897, "eval_ORGANIZATION_recall": 0.8533333333333334, "eval_PERSON_f1": 0.8859060402684563, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8516129032258064, "eval_PERSON_recall": 0.9230769230769231, "eval_QUANTITY_f1": 0.6415094339622641, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5666666666666667, "eval_QUANTITY_recall": 0.7391304347826086, "eval_TIME_f1": 0.7796610169491526, "eval_TIME_number": 26, "eval_TIME_precision": 0.696969696969697, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.16515356302261353, "eval_overall_accuracy": 0.9564673505128847, "eval_overall_f1": 0.8099173553719009, "eval_overall_precision": 0.7596899224806202, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.3795, "eval_samples_per_second": 492.77, "eval_steps_per_second": 7.905, "step": 3498 }, { "epoch": 34.0, "grad_norm": 0.9991981387138367, "learning_rate": 3.3e-05, "loss": 0.0588, "step": 3604 }, { "epoch": 34.0, "eval_LOCATION_f1": 0.8137931034482759, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8082191780821918, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.7692307692307693, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6914893617021277, "eval_ORGANIZATION_recall": 0.8666666666666667, "eval_PERSON_f1": 0.888888888888889, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8571428571428571, "eval_PERSON_recall": 0.9230769230769231, "eval_QUANTITY_f1": 0.6923076923076923, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6206896551724138, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.7586206896551724, "eval_TIME_number": 26, "eval_TIME_precision": 0.6875, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.1578424870967865, "eval_overall_accuracy": 0.9592194145609206, "eval_overall_f1": 0.8210818307905686, "eval_overall_precision": 0.774869109947644, "eval_overall_recall": 0.8731563421828908, "eval_runtime": 0.3816, "eval_samples_per_second": 490.007, "eval_steps_per_second": 7.861, "step": 3604 }, { "epoch": 35.0, "grad_norm": 0.7979607582092285, "learning_rate": 3.2500000000000004e-05, "loss": 0.0592, "step": 3710 }, { "epoch": 35.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7398843930635838, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6530612244897959, "eval_ORGANIZATION_recall": 0.8533333333333334, "eval_PERSON_f1": 0.8821548821548821, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8506493506493507, "eval_PERSON_recall": 0.916083916083916, "eval_QUANTITY_f1": 0.6792452830188679, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.7586206896551724, "eval_TIME_number": 26, "eval_TIME_precision": 0.6875, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.16431660950183868, "eval_overall_accuracy": 0.9552164123092319, "eval_overall_f1": 0.8104395604395604, "eval_overall_precision": 0.7583547557840618, "eval_overall_recall": 0.8702064896755162, "eval_runtime": 0.3788, "eval_samples_per_second": 493.707, "eval_steps_per_second": 7.92, "step": 3710 }, { "epoch": 36.0, "grad_norm": 0.592914342880249, "learning_rate": 3.2000000000000005e-05, "loss": 0.0587, "step": 3816 }, { "epoch": 36.0, "eval_LOCATION_f1": 0.7947019867549668, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.759493670886076, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.75, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6774193548387096, "eval_ORGANIZATION_recall": 0.84, "eval_PERSON_f1": 0.8986486486486487, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.869281045751634, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.6792452830188679, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8214285714285715, "eval_TIME_number": 26, "eval_TIME_precision": 0.7666666666666667, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.15960274636745453, "eval_overall_accuracy": 0.9594696022016512, "eval_overall_f1": 0.8204419889502762, "eval_overall_precision": 0.7714285714285715, "eval_overall_recall": 0.8761061946902655, "eval_runtime": 0.3801, "eval_samples_per_second": 492.012, "eval_steps_per_second": 7.893, "step": 3816 }, { "epoch": 37.0, "grad_norm": 1.3007019758224487, "learning_rate": 3.15e-05, "loss": 0.0554, "step": 3922 }, { "epoch": 37.0, "eval_LOCATION_f1": 0.8169014084507044, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8285714285714286, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.7469879518072289, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6813186813186813, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.888888888888889, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8571428571428571, "eval_PERSON_recall": 0.9230769230769231, "eval_QUANTITY_f1": 0.7058823529411765, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6428571428571429, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.7777777777777779, "eval_TIME_number": 26, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.15752674639225006, "eval_overall_accuracy": 0.9594696022016512, "eval_overall_f1": 0.819718309859155, "eval_overall_precision": 0.784366576819407, "eval_overall_recall": 0.8584070796460177, "eval_runtime": 0.3795, "eval_samples_per_second": 492.698, "eval_steps_per_second": 7.904, "step": 3922 }, { "epoch": 38.0, "grad_norm": 0.44553250074386597, "learning_rate": 3.1e-05, "loss": 0.0552, "step": 4028 }, { "epoch": 38.0, "eval_LOCATION_f1": 0.8055555555555556, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8055555555555556, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.751592356687898, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7195121951219512, "eval_ORGANIZATION_recall": 0.7866666666666666, "eval_PERSON_f1": 0.9023569023569024, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8701298701298701, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.6923076923076923, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6206896551724138, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.7931034482758621, "eval_TIME_number": 26, "eval_TIME_precision": 0.71875, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.15892784297466278, "eval_overall_accuracy": 0.961220915686765, "eval_overall_f1": 0.8248587570621468, "eval_overall_precision": 0.7913279132791328, "eval_overall_recall": 0.8613569321533924, "eval_runtime": 0.3796, "eval_samples_per_second": 492.603, "eval_steps_per_second": 7.903, "step": 4028 }, { "epoch": 39.0, "grad_norm": 1.0582730770111084, "learning_rate": 3.05e-05, "loss": 0.0541, "step": 4134 }, { "epoch": 39.0, "eval_LOCATION_f1": 0.7916666666666666, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7916666666666666, "eval_LOCATION_recall": 0.7916666666666666, "eval_ORGANIZATION_f1": 0.7349397590361445, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6703296703296703, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.8837209302325583, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8417721518987342, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.6666666666666667, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.5806451612903226, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.7368421052631579, "eval_TIME_number": 26, "eval_TIME_precision": 0.6774193548387096, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.16924403607845306, "eval_overall_accuracy": 0.9559669752314236, "eval_overall_f1": 0.8033240997229917, "eval_overall_precision": 0.7571801566579635, "eval_overall_recall": 0.855457227138643, "eval_runtime": 0.3796, "eval_samples_per_second": 492.605, "eval_steps_per_second": 7.903, "step": 4134 }, { "epoch": 40.0, "grad_norm": 0.699941873550415, "learning_rate": 3e-05, "loss": 0.0509, "step": 4240 }, { "epoch": 40.0, "eval_LOCATION_f1": 0.7891156462585034, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7733333333333333, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.751592356687898, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7195121951219512, "eval_ORGANIZATION_recall": 0.7866666666666666, "eval_PERSON_f1": 0.8963210702341137, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8589743589743589, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7346938775510203, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6923076923076923, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8275862068965517, "eval_TIME_number": 26, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.9230769230769231, "eval_loss": 0.16636206209659576, "eval_overall_accuracy": 0.9597197898423818, "eval_overall_f1": 0.8253521126760562, "eval_overall_precision": 0.7897574123989218, "eval_overall_recall": 0.8643067846607669, "eval_runtime": 0.3789, "eval_samples_per_second": 493.489, "eval_steps_per_second": 7.917, "step": 4240 }, { "epoch": 41.0, "grad_norm": 1.5197551250457764, "learning_rate": 2.95e-05, "loss": 0.0519, "step": 4346 }, { "epoch": 41.0, "eval_LOCATION_f1": 0.8275862068965517, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.821917808219178, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7471264367816093, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6565656565656566, "eval_ORGANIZATION_recall": 0.8666666666666667, "eval_PERSON_f1": 0.8933333333333333, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8535031847133758, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.6923076923076923, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6206896551724138, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8275862068965517, "eval_TIME_number": 26, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.9230769230769231, "eval_loss": 0.174987331032753, "eval_overall_accuracy": 0.9569677257943457, "eval_overall_f1": 0.8257887517146778, "eval_overall_precision": 0.7717948717948718, "eval_overall_recall": 0.887905604719764, "eval_runtime": 0.3783, "eval_samples_per_second": 494.313, "eval_steps_per_second": 7.93, "step": 4346 }, { "epoch": 42.0, "grad_norm": 1.1461714506149292, "learning_rate": 2.9e-05, "loss": 0.053, "step": 4452 }, { "epoch": 42.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.782051282051282, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7530864197530864, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.8821548821548821, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8506493506493507, "eval_PERSON_recall": 0.916083916083916, "eval_QUANTITY_f1": 0.6792452830188679, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8070175438596492, "eval_TIME_number": 26, "eval_TIME_precision": 0.7419354838709677, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.16791585087776184, "eval_overall_accuracy": 0.9582186639979985, "eval_overall_f1": 0.8253521126760562, "eval_overall_precision": 0.7897574123989218, "eval_overall_recall": 0.8643067846607669, "eval_runtime": 0.3802, "eval_samples_per_second": 491.843, "eval_steps_per_second": 7.891, "step": 4452 }, { "epoch": 43.0, "grad_norm": 0.24822089076042175, "learning_rate": 2.8499999999999998e-05, "loss": 0.0497, "step": 4558 }, { "epoch": 43.0, "eval_LOCATION_f1": 0.8243243243243243, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8026315789473685, "eval_LOCATION_recall": 0.8472222222222222, "eval_ORGANIZATION_f1": 0.7581699346405228, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7435897435897436, "eval_ORGANIZATION_recall": 0.7733333333333333, "eval_PERSON_f1": 0.8963210702341137, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8589743589743589, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.6792452830188679, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8363636363636363, "eval_TIME_number": 26, "eval_TIME_precision": 0.7931034482758621, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.16703499853610992, "eval_overall_accuracy": 0.9594696022016512, "eval_overall_f1": 0.8305084745762712, "eval_overall_precision": 0.7967479674796748, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.3792, "eval_samples_per_second": 493.197, "eval_steps_per_second": 7.912, "step": 4558 }, { "epoch": 44.0, "grad_norm": 1.3078693151474, "learning_rate": 2.8000000000000003e-05, "loss": 0.0496, "step": 4664 }, { "epoch": 44.0, "eval_LOCATION_f1": 0.7919463087248323, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7662337662337663, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.7654320987654321, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7126436781609196, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7755102040816326, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7307692307692307, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.16706283390522003, "eval_overall_accuracy": 0.9577182887165374, "eval_overall_f1": 0.8263305322128851, "eval_overall_precision": 0.7866666666666666, "eval_overall_recall": 0.8702064896755162, "eval_runtime": 0.3771, "eval_samples_per_second": 495.916, "eval_steps_per_second": 7.956, "step": 4664 }, { "epoch": 45.0, "grad_norm": 0.7571492195129395, "learning_rate": 2.7500000000000004e-05, "loss": 0.0499, "step": 4770 }, { "epoch": 45.0, "eval_LOCATION_f1": 0.7919463087248323, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7662337662337663, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.7320261437908497, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.717948717948718, "eval_ORGANIZATION_recall": 0.7466666666666667, "eval_PERSON_f1": 0.8741721854304637, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8301886792452831, "eval_PERSON_recall": 0.9230769230769231, "eval_QUANTITY_f1": 0.6923076923076923, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6206896551724138, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.7857142857142856, "eval_TIME_number": 26, "eval_TIME_precision": 0.7333333333333333, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.18108969926834106, "eval_overall_accuracy": 0.9559669752314236, "eval_overall_f1": 0.8061797752808988, "eval_overall_precision": 0.7694369973190348, "eval_overall_recall": 0.8466076696165191, "eval_runtime": 0.3799, "eval_samples_per_second": 492.21, "eval_steps_per_second": 7.896, "step": 4770 }, { "epoch": 46.0, "grad_norm": 0.27301469445228577, "learning_rate": 2.7000000000000002e-05, "loss": 0.0457, "step": 4876 }, { "epoch": 46.0, "eval_LOCATION_f1": 0.8108108108108109, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7894736842105263, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7547169811320756, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7142857142857143, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8963210702341137, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8589743589743589, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7169811320754716, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6333333333333333, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8214285714285715, "eval_TIME_number": 26, "eval_TIME_precision": 0.7666666666666667, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.1713663935661316, "eval_overall_accuracy": 0.9592194145609206, "eval_overall_f1": 0.827972027972028, "eval_overall_precision": 0.7872340425531915, "eval_overall_recall": 0.8731563421828908, "eval_runtime": 0.3794, "eval_samples_per_second": 492.839, "eval_steps_per_second": 7.907, "step": 4876 }, { "epoch": 47.0, "grad_norm": 0.6634193658828735, "learning_rate": 2.6500000000000004e-05, "loss": 0.0479, "step": 4982 }, { "epoch": 47.0, "eval_LOCATION_f1": 0.8, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7945205479452054, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.7411764705882353, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6631578947368421, "eval_ORGANIZATION_recall": 0.84, "eval_PERSON_f1": 0.8896321070234114, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8525641025641025, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7037037037037037, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6129032258064516, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8363636363636363, "eval_TIME_number": 26, "eval_TIME_precision": 0.7931034482758621, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.1743399053812027, "eval_overall_accuracy": 0.9569677257943457, "eval_overall_f1": 0.8188105117565698, "eval_overall_precision": 0.7708333333333334, "eval_overall_recall": 0.8731563421828908, "eval_runtime": 0.38, "eval_samples_per_second": 492.071, "eval_steps_per_second": 7.894, "step": 4982 }, { "epoch": 48.0, "grad_norm": 0.546776294708252, "learning_rate": 2.6000000000000002e-05, "loss": 0.0451, "step": 5088 }, { "epoch": 48.0, "eval_LOCATION_f1": 0.7945205479452055, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7837837837837838, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.7721518987341773, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7349397590361446, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.8986486486486487, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.869281045751634, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8363636363636363, "eval_TIME_number": 26, "eval_TIME_precision": 0.7931034482758621, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.17288771271705627, "eval_overall_accuracy": 0.9602201651238429, "eval_overall_f1": 0.8340425531914893, "eval_overall_precision": 0.8032786885245902, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.3801, "eval_samples_per_second": 491.927, "eval_steps_per_second": 7.892, "step": 5088 }, { "epoch": 49.0, "grad_norm": 0.645470142364502, "learning_rate": 2.5500000000000003e-05, "loss": 0.0456, "step": 5194 }, { "epoch": 49.0, "eval_LOCATION_f1": 0.8053691275167787, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7792207792207793, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7848101265822784, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7469879518072289, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.8926174496644295, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8580645161290322, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8070175438596492, "eval_TIME_number": 26, "eval_TIME_precision": 0.7419354838709677, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.1743689626455307, "eval_overall_accuracy": 0.9592194145609206, "eval_overall_f1": 0.8330995792426367, "eval_overall_precision": 0.7941176470588235, "eval_overall_recall": 0.8761061946902655, "eval_runtime": 0.3793, "eval_samples_per_second": 493.031, "eval_steps_per_second": 7.91, "step": 5194 }, { "epoch": 50.0, "grad_norm": 0.8035615682601929, "learning_rate": 2.5e-05, "loss": 0.0422, "step": 5300 }, { "epoch": 50.0, "eval_LOCATION_f1": 0.8053691275167787, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7792207792207793, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.728395061728395, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6781609195402298, "eval_ORGANIZATION_recall": 0.7866666666666666, "eval_PERSON_f1": 0.9023569023569024, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8701298701298701, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.6923076923076923, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6206896551724138, "eval_QUANTITY_recall": 0.782608695652174, "eval_TIME_f1": 0.8148148148148148, "eval_TIME_number": 26, "eval_TIME_precision": 0.7857142857142857, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.17739371955394745, "eval_overall_accuracy": 0.9584688516387291, "eval_overall_f1": 0.8207282913165266, "eval_overall_precision": 0.7813333333333333, "eval_overall_recall": 0.8643067846607669, "eval_runtime": 0.3805, "eval_samples_per_second": 491.473, "eval_steps_per_second": 7.885, "step": 5300 }, { "epoch": 51.0, "grad_norm": 0.6081034541130066, "learning_rate": 2.45e-05, "loss": 0.0435, "step": 5406 }, { "epoch": 51.0, "eval_LOCATION_f1": 0.8133333333333332, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.782051282051282, "eval_LOCATION_recall": 0.8472222222222222, "eval_ORGANIZATION_f1": 0.785276073619632, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7272727272727273, "eval_ORGANIZATION_recall": 0.8533333333333334, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7307692307692308, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6551724137931034, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7857142857142856, "eval_TIME_number": 26, "eval_TIME_precision": 0.7333333333333333, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.17463435232639313, "eval_overall_accuracy": 0.9597197898423818, "eval_overall_f1": 0.8328690807799444, "eval_overall_precision": 0.7889182058047494, "eval_overall_recall": 0.8820058997050148, "eval_runtime": 0.3825, "eval_samples_per_second": 488.888, "eval_steps_per_second": 7.843, "step": 5406 }, { "epoch": 52.0, "grad_norm": 2.0089409351348877, "learning_rate": 2.4e-05, "loss": 0.0439, "step": 5512 }, { "epoch": 52.0, "eval_LOCATION_f1": 0.7972972972972973, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7763157894736842, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.7425149700598803, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6739130434782609, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.8986486486486487, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.869281045751634, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8070175438596492, "eval_TIME_number": 26, "eval_TIME_precision": 0.7419354838709677, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.17734888195991516, "eval_overall_accuracy": 0.9602201651238429, "eval_overall_f1": 0.8245125348189415, "eval_overall_precision": 0.7810026385224275, "eval_overall_recall": 0.8731563421828908, "eval_runtime": 0.3818, "eval_samples_per_second": 489.752, "eval_steps_per_second": 7.857, "step": 5512 }, { "epoch": 53.0, "grad_norm": 0.5437747240066528, "learning_rate": 2.35e-05, "loss": 0.0416, "step": 5618 }, { "epoch": 53.0, "eval_LOCATION_f1": 0.802721088435374, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7866666666666666, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.7349397590361445, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6703296703296703, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.9054054054054055, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8758169934640523, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7169811320754716, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6333333333333333, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8518518518518519, "eval_TIME_number": 26, "eval_TIME_precision": 0.8214285714285714, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.17923349142074585, "eval_overall_accuracy": 0.9594696022016512, "eval_overall_f1": 0.8268156424581006, "eval_overall_precision": 0.7851458885941645, "eval_overall_recall": 0.8731563421828908, "eval_runtime": 0.3808, "eval_samples_per_second": 491.045, "eval_steps_per_second": 7.878, "step": 5618 }, { "epoch": 54.0, "grad_norm": 0.4538353681564331, "learning_rate": 2.3000000000000003e-05, "loss": 0.0399, "step": 5724 }, { "epoch": 54.0, "eval_LOCATION_f1": 0.8053691275167787, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7792207792207793, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7730061349693251, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7159090909090909, "eval_ORGANIZATION_recall": 0.84, "eval_PERSON_f1": 0.9023569023569024, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8701298701298701, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7307692307692308, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6551724137931034, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.75, "eval_TIME_number": 26, "eval_TIME_precision": 0.7, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.1828322857618332, "eval_overall_accuracy": 0.9589692269201902, "eval_overall_f1": 0.8284518828451883, "eval_overall_precision": 0.7857142857142857, "eval_overall_recall": 0.8761061946902655, "eval_runtime": 0.3807, "eval_samples_per_second": 491.24, "eval_steps_per_second": 7.881, "step": 5724 }, { "epoch": 55.0, "grad_norm": 0.803385853767395, "learning_rate": 2.25e-05, "loss": 0.0422, "step": 5830 }, { "epoch": 55.0, "eval_LOCATION_f1": 0.8053691275167787, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7792207792207793, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7558139534883721, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6701030927835051, "eval_ORGANIZATION_recall": 0.8666666666666667, "eval_PERSON_f1": 0.9054054054054055, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8758169934640523, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7857142857142856, "eval_TIME_number": 26, "eval_TIME_precision": 0.7333333333333333, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.18513092398643494, "eval_overall_accuracy": 0.9587190392794596, "eval_overall_f1": 0.8298755186721991, "eval_overall_precision": 0.78125, "eval_overall_recall": 0.8849557522123894, "eval_runtime": 0.381, "eval_samples_per_second": 490.868, "eval_steps_per_second": 7.875, "step": 5830 }, { "epoch": 56.0, "grad_norm": 0.6640517115592957, "learning_rate": 2.2000000000000003e-05, "loss": 0.0426, "step": 5936 }, { "epoch": 56.0, "eval_LOCATION_f1": 0.7891156462585034, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7733333333333333, "eval_LOCATION_recall": 0.8055555555555556, "eval_ORGANIZATION_f1": 0.7770700636942677, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7439024390243902, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.9023569023569024, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8701298701298701, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8, "eval_TIME_number": 26, "eval_TIME_precision": 0.7586206896551724, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.1791711300611496, "eval_overall_accuracy": 0.961220915686765, "eval_overall_f1": 0.8316831683168316, "eval_overall_precision": 0.7989130434782609, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.3792, "eval_samples_per_second": 493.183, "eval_steps_per_second": 7.912, "step": 5936 }, { "epoch": 57.0, "grad_norm": 0.5826635360717773, "learning_rate": 2.15e-05, "loss": 0.0414, "step": 6042 }, { "epoch": 57.0, "eval_LOCATION_f1": 0.802721088435374, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7866666666666666, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.782051282051282, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7530864197530864, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.888888888888889, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8571428571428571, "eval_PERSON_recall": 0.9230769230769231, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8076923076923077, "eval_TIME_number": 26, "eval_TIME_precision": 0.8076923076923077, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.17410559952259064, "eval_overall_accuracy": 0.9609707280460346, "eval_overall_f1": 0.8307254623044097, "eval_overall_precision": 0.8021978021978022, "eval_overall_recall": 0.8613569321533924, "eval_runtime": 0.382, "eval_samples_per_second": 489.466, "eval_steps_per_second": 7.852, "step": 6042 }, { "epoch": 58.0, "grad_norm": 0.5141724944114685, "learning_rate": 2.1e-05, "loss": 0.0386, "step": 6148 }, { "epoch": 58.0, "eval_LOCATION_f1": 0.8266666666666667, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7948717948717948, "eval_LOCATION_recall": 0.8611111111111112, "eval_ORGANIZATION_f1": 0.785276073619632, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7272727272727273, "eval_ORGANIZATION_recall": 0.8533333333333334, "eval_PERSON_f1": 0.8986486486486487, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.869281045751634, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7037037037037037, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6129032258064516, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7777777777777779, "eval_TIME_number": 26, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.1851791888475418, "eval_overall_accuracy": 0.960720540405304, "eval_overall_f1": 0.8340306834030684, "eval_overall_precision": 0.791005291005291, "eval_overall_recall": 0.8820058997050148, "eval_runtime": 0.383, "eval_samples_per_second": 488.276, "eval_steps_per_second": 7.833, "step": 6148 }, { "epoch": 59.0, "grad_norm": 0.39003196358680725, "learning_rate": 2.05e-05, "loss": 0.0391, "step": 6254 }, { "epoch": 59.0, "eval_LOCATION_f1": 0.8026315789473685, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7625, "eval_LOCATION_recall": 0.8472222222222222, "eval_ORGANIZATION_f1": 0.7636363636363636, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7, "eval_ORGANIZATION_recall": 0.84, "eval_PERSON_f1": 0.9054054054054055, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8758169934640523, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7307692307692308, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6551724137931034, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7241379310344829, "eval_TIME_number": 26, "eval_TIME_precision": 0.65625, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.18867133557796478, "eval_overall_accuracy": 0.9584688516387291, "eval_overall_f1": 0.8243430152143844, "eval_overall_precision": 0.7760416666666666, "eval_overall_recall": 0.8790560471976401, "eval_runtime": 0.3802, "eval_samples_per_second": 491.903, "eval_steps_per_second": 7.891, "step": 6254 }, { "epoch": 60.0, "grad_norm": 0.5746458172798157, "learning_rate": 2e-05, "loss": 0.0377, "step": 6360 }, { "epoch": 60.0, "eval_LOCATION_f1": 0.8079470198675496, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7721518987341772, "eval_LOCATION_recall": 0.8472222222222222, "eval_ORGANIZATION_f1": 0.7710843373493976, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7032967032967034, "eval_ORGANIZATION_recall": 0.8533333333333334, "eval_PERSON_f1": 0.8993288590604027, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.864516129032258, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7037037037037037, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6129032258064516, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.75, "eval_TIME_number": 26, "eval_TIME_precision": 0.7, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.1903800368309021, "eval_overall_accuracy": 0.9574681010758068, "eval_overall_f1": 0.8248275862068966, "eval_overall_precision": 0.7746113989637305, "eval_overall_recall": 0.8820058997050148, "eval_runtime": 0.3787, "eval_samples_per_second": 493.746, "eval_steps_per_second": 7.921, "step": 6360 }, { "epoch": 61.0, "grad_norm": 1.1122630834579468, "learning_rate": 1.9500000000000003e-05, "loss": 0.0377, "step": 6466 }, { "epoch": 61.0, "eval_LOCATION_f1": 0.8082191780821918, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7972972972972973, "eval_LOCATION_recall": 0.8194444444444444, "eval_ORGANIZATION_f1": 0.7407407407407408, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6896551724137931, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7307692307692308, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6551724137931034, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8, "eval_TIME_number": 26, "eval_TIME_precision": 0.7586206896551724, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.19008861482143402, "eval_overall_accuracy": 0.957968476357268, "eval_overall_f1": 0.8230337078651685, "eval_overall_precision": 0.7855227882037533, "eval_overall_recall": 0.8643067846607669, "eval_runtime": 0.3825, "eval_samples_per_second": 488.926, "eval_steps_per_second": 7.844, "step": 6466 }, { "epoch": 62.0, "grad_norm": 0.9635010957717896, "learning_rate": 1.9e-05, "loss": 0.0365, "step": 6572 }, { "epoch": 62.0, "eval_LOCATION_f1": 0.8108108108108109, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7894736842105263, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7590361445783131, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6923076923076923, "eval_ORGANIZATION_recall": 0.84, "eval_PERSON_f1": 0.9023569023569024, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8701298701298701, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7169811320754716, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6333333333333333, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.75, "eval_TIME_number": 26, "eval_TIME_precision": 0.7, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.19233578443527222, "eval_overall_accuracy": 0.9584688516387291, "eval_overall_f1": 0.825, "eval_overall_precision": 0.7795275590551181, "eval_overall_recall": 0.8761061946902655, "eval_runtime": 0.3825, "eval_samples_per_second": 488.872, "eval_steps_per_second": 7.843, "step": 6572 }, { "epoch": 63.0, "grad_norm": 0.4081244170665741, "learning_rate": 1.85e-05, "loss": 0.0381, "step": 6678 }, { "epoch": 63.0, "eval_LOCATION_f1": 0.8435374149659863, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8266666666666667, "eval_LOCATION_recall": 0.8611111111111112, "eval_ORGANIZATION_f1": 0.7607361963190183, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7045454545454546, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.8926174496644295, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8580645161290322, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7777777777777779, "eval_TIME_number": 26, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.18928496539592743, "eval_overall_accuracy": 0.9599699774831123, "eval_overall_f1": 0.8330995792426367, "eval_overall_precision": 0.7941176470588235, "eval_overall_recall": 0.8761061946902655, "eval_runtime": 0.3837, "eval_samples_per_second": 487.415, "eval_steps_per_second": 7.819, "step": 6678 }, { "epoch": 64.0, "grad_norm": 0.38579753041267395, "learning_rate": 1.8e-05, "loss": 0.0356, "step": 6784 }, { "epoch": 64.0, "eval_LOCATION_f1": 0.8378378378378377, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8157894736842105, "eval_LOCATION_recall": 0.8611111111111112, "eval_ORGANIZATION_f1": 0.7560975609756098, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6966292134831461, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.8986486486486487, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.869281045751634, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.75, "eval_TIME_number": 26, "eval_TIME_precision": 0.7, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.1920894831418991, "eval_overall_accuracy": 0.9592194145609206, "eval_overall_f1": 0.8307692307692307, "eval_overall_precision": 0.7898936170212766, "eval_overall_recall": 0.8761061946902655, "eval_runtime": 0.3803, "eval_samples_per_second": 491.656, "eval_steps_per_second": 7.888, "step": 6784 }, { "epoch": 65.0, "grad_norm": 0.6564523577690125, "learning_rate": 1.75e-05, "loss": 0.0359, "step": 6890 }, { "epoch": 65.0, "eval_LOCATION_f1": 0.8187919463087249, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7922077922077922, "eval_LOCATION_recall": 0.8472222222222222, "eval_ORGANIZATION_f1": 0.7672955974842767, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7261904761904762, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.8754208754208753, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8441558441558441, "eval_PERSON_recall": 0.9090909090909091, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8, "eval_TIME_number": 26, "eval_TIME_precision": 0.7586206896551724, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.1849316507577896, "eval_overall_accuracy": 0.9597197898423818, "eval_overall_f1": 0.8241912798874824, "eval_overall_precision": 0.7876344086021505, "eval_overall_recall": 0.8643067846607669, "eval_runtime": 0.3795, "eval_samples_per_second": 492.748, "eval_steps_per_second": 7.905, "step": 6890 }, { "epoch": 66.0, "grad_norm": 0.21363137662410736, "learning_rate": 1.7000000000000003e-05, "loss": 0.0352, "step": 6996 }, { "epoch": 66.0, "eval_LOCATION_f1": 0.8243243243243243, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8026315789473685, "eval_LOCATION_recall": 0.8472222222222222, "eval_ORGANIZATION_f1": 0.7607361963190183, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7045454545454546, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.8926174496644295, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8580645161290322, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7307692307692308, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6551724137931034, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8, "eval_TIME_number": 26, "eval_TIME_precision": 0.7586206896551724, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.1957569569349289, "eval_overall_accuracy": 0.9597197898423818, "eval_overall_f1": 0.8296089385474861, "eval_overall_precision": 0.7877984084880637, "eval_overall_recall": 0.8761061946902655, "eval_runtime": 0.3828, "eval_samples_per_second": 488.457, "eval_steps_per_second": 7.836, "step": 6996 }, { "epoch": 67.0, "grad_norm": 0.11434553563594818, "learning_rate": 1.65e-05, "loss": 0.0339, "step": 7102 }, { "epoch": 67.0, "eval_LOCATION_f1": 0.8299319727891157, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8133333333333334, "eval_LOCATION_recall": 0.8472222222222222, "eval_ORGANIZATION_f1": 0.7701863354037267, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7209302325581395, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.8986486486486487, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.869281045751634, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7307692307692308, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6551724137931034, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7777777777777779, "eval_TIME_number": 26, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.1901414692401886, "eval_overall_accuracy": 0.9609707280460346, "eval_overall_f1": 0.8338028169014085, "eval_overall_precision": 0.7978436657681941, "eval_overall_recall": 0.8731563421828908, "eval_runtime": 0.3815, "eval_samples_per_second": 490.135, "eval_steps_per_second": 7.863, "step": 7102 }, { "epoch": 68.0, "grad_norm": 0.28407835960388184, "learning_rate": 1.6000000000000003e-05, "loss": 0.0349, "step": 7208 }, { "epoch": 68.0, "eval_LOCATION_f1": 0.823529411764706, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7777777777777778, "eval_LOCATION_recall": 0.875, "eval_ORGANIZATION_f1": 0.75, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6774193548387096, "eval_ORGANIZATION_recall": 0.84, "eval_PERSON_f1": 0.9023569023569024, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8701298701298701, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7017543859649122, "eval_TIME_number": 26, "eval_TIME_precision": 0.6451612903225806, "eval_TIME_recall": 0.7692307692307693, "eval_loss": 0.20368348062038422, "eval_overall_accuracy": 0.9562171628721541, "eval_overall_f1": 0.8236914600550964, "eval_overall_precision": 0.772609819121447, "eval_overall_recall": 0.8820058997050148, "eval_runtime": 0.378, "eval_samples_per_second": 494.725, "eval_steps_per_second": 7.937, "step": 7208 }, { "epoch": 69.0, "grad_norm": 0.4101424515247345, "learning_rate": 1.55e-05, "loss": 0.0338, "step": 7314 }, { "epoch": 69.0, "eval_LOCATION_f1": 0.8219178082191781, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8108108108108109, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7453416149068322, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6976744186046512, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8963210702341137, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8589743589743589, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7924528301886792, "eval_TIME_number": 26, "eval_TIME_precision": 0.7777777777777778, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.19146528840065002, "eval_overall_accuracy": 0.9594696022016512, "eval_overall_f1": 0.8293370944992948, "eval_overall_precision": 0.7945945945945946, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.3851, "eval_samples_per_second": 485.558, "eval_steps_per_second": 7.79, "step": 7314 }, { "epoch": 70.0, "grad_norm": 0.7535924911499023, "learning_rate": 1.5e-05, "loss": 0.0343, "step": 7420 }, { "epoch": 70.0, "eval_LOCATION_f1": 0.8275862068965517, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.821917808219178, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7692307692307692, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7407407407407407, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8993288590604027, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.864516129032258, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7636363636363636, "eval_TIME_number": 26, "eval_TIME_precision": 0.7241379310344828, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.18880921602249146, "eval_overall_accuracy": 0.9597197898423818, "eval_overall_f1": 0.8352272727272728, "eval_overall_precision": 0.8054794520547945, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.3792, "eval_samples_per_second": 493.182, "eval_steps_per_second": 7.912, "step": 7420 }, { "epoch": 71.0, "grad_norm": 0.4689802825450897, "learning_rate": 1.45e-05, "loss": 0.0328, "step": 7526 }, { "epoch": 71.0, "eval_LOCATION_f1": 0.8219178082191781, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8108108108108109, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7607361963190183, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7045454545454546, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.8949152542372882, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.868421052631579, "eval_PERSON_recall": 0.9230769230769231, "eval_QUANTITY_f1": 0.7755102040816326, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7307692307692307, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8, "eval_TIME_number": 26, "eval_TIME_precision": 0.7586206896551724, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.18770109117031097, "eval_overall_accuracy": 0.9599699774831123, "eval_overall_f1": 0.8333333333333334, "eval_overall_precision": 0.7994579945799458, "eval_overall_recall": 0.8702064896755162, "eval_runtime": 0.3827, "eval_samples_per_second": 488.68, "eval_steps_per_second": 7.84, "step": 7526 }, { "epoch": 72.0, "grad_norm": 0.7137999534606934, "learning_rate": 1.4000000000000001e-05, "loss": 0.0323, "step": 7632 }, { "epoch": 72.0, "eval_LOCATION_f1": 0.8219178082191781, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8108108108108109, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7421383647798742, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7023809523809523, "eval_ORGANIZATION_recall": 0.7866666666666666, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7755102040816326, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7307692307692307, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.19658206403255463, "eval_overall_accuracy": 0.9597197898423818, "eval_overall_f1": 0.8276836158192091, "eval_overall_precision": 0.7940379403794038, "eval_overall_recall": 0.8643067846607669, "eval_runtime": 0.3812, "eval_samples_per_second": 490.568, "eval_steps_per_second": 7.87, "step": 7632 }, { "epoch": 73.0, "grad_norm": 0.905379056930542, "learning_rate": 1.3500000000000001e-05, "loss": 0.0322, "step": 7738 }, { "epoch": 73.0, "eval_LOCATION_f1": 0.8219178082191781, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8108108108108109, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7701863354037267, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7209302325581395, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.9090909090909092, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8766233766233766, "eval_PERSON_recall": 0.9440559440559441, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8148148148148148, "eval_TIME_number": 26, "eval_TIME_precision": 0.7857142857142857, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.1963140368461609, "eval_overall_accuracy": 0.9602201651238429, "eval_overall_f1": 0.8418079096045198, "eval_overall_precision": 0.8075880758807588, "eval_overall_recall": 0.8790560471976401, "eval_runtime": 0.3793, "eval_samples_per_second": 493.015, "eval_steps_per_second": 7.909, "step": 7738 }, { "epoch": 74.0, "grad_norm": 1.4441778659820557, "learning_rate": 1.3000000000000001e-05, "loss": 0.0315, "step": 7844 }, { "epoch": 74.0, "eval_LOCATION_f1": 0.8243243243243243, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8026315789473685, "eval_LOCATION_recall": 0.8472222222222222, "eval_ORGANIZATION_f1": 0.7500000000000001, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7058823529411765, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.9090909090909092, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8766233766233766, "eval_PERSON_recall": 0.9440559440559441, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.19720590114593506, "eval_overall_accuracy": 0.9594696022016512, "eval_overall_f1": 0.8330995792426367, "eval_overall_precision": 0.7941176470588235, "eval_overall_recall": 0.8761061946902655, "eval_runtime": 0.3807, "eval_samples_per_second": 491.142, "eval_steps_per_second": 7.879, "step": 7844 }, { "epoch": 75.0, "grad_norm": 0.8844290971755981, "learning_rate": 1.25e-05, "loss": 0.032, "step": 7950 }, { "epoch": 75.0, "eval_LOCATION_f1": 0.8108108108108109, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7894736842105263, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7547169811320756, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7142857142857143, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8993288590604027, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.864516129032258, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8, "eval_TIME_number": 26, "eval_TIME_precision": 0.7586206896551724, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.20169958472251892, "eval_overall_accuracy": 0.9597197898423818, "eval_overall_f1": 0.829817158931083, "eval_overall_precision": 0.793010752688172, "eval_overall_recall": 0.8702064896755162, "eval_runtime": 0.3914, "eval_samples_per_second": 477.748, "eval_steps_per_second": 7.664, "step": 7950 }, { "epoch": 76.0, "grad_norm": 1.7479166984558105, "learning_rate": 1.2e-05, "loss": 0.03, "step": 8056 }, { "epoch": 76.0, "eval_LOCATION_f1": 0.8108108108108109, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7894736842105263, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7784431137724551, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7065217391304348, "eval_ORGANIZATION_recall": 0.8666666666666667, "eval_PERSON_f1": 0.8926174496644295, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8580645161290322, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8, "eval_TIME_number": 26, "eval_TIME_precision": 0.7586206896551724, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.2011711448431015, "eval_overall_accuracy": 0.9597197898423818, "eval_overall_f1": 0.8328690807799444, "eval_overall_precision": 0.7889182058047494, "eval_overall_recall": 0.8820058997050148, "eval_runtime": 0.3816, "eval_samples_per_second": 490.049, "eval_steps_per_second": 7.862, "step": 8056 }, { "epoch": 77.0, "grad_norm": 0.3027796447277069, "learning_rate": 1.1500000000000002e-05, "loss": 0.0319, "step": 8162 }, { "epoch": 77.0, "eval_LOCATION_f1": 0.8299319727891157, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8133333333333334, "eval_LOCATION_recall": 0.8472222222222222, "eval_ORGANIZATION_f1": 0.7453416149068322, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6976744186046512, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8963210702341137, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8589743589743589, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8, "eval_TIME_number": 26, "eval_TIME_precision": 0.7586206896551724, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.1984999179840088, "eval_overall_accuracy": 0.9592194145609206, "eval_overall_f1": 0.8314606741573034, "eval_overall_precision": 0.7935656836461126, "eval_overall_recall": 0.8731563421828908, "eval_runtime": 0.3788, "eval_samples_per_second": 493.685, "eval_steps_per_second": 7.92, "step": 8162 }, { "epoch": 78.0, "grad_norm": 0.4953755736351013, "learning_rate": 1.1000000000000001e-05, "loss": 0.0291, "step": 8268 }, { "epoch": 78.0, "eval_LOCATION_f1": 0.8187919463087249, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7922077922077922, "eval_LOCATION_recall": 0.8472222222222222, "eval_ORGANIZATION_f1": 0.7484662576687117, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6931818181818182, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.9054054054054055, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8758169934640523, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8, "eval_TIME_number": 26, "eval_TIME_precision": 0.7586206896551724, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.20313413441181183, "eval_overall_accuracy": 0.9599699774831123, "eval_overall_f1": 0.8319327731092437, "eval_overall_precision": 0.792, "eval_overall_recall": 0.8761061946902655, "eval_runtime": 0.3815, "eval_samples_per_second": 490.185, "eval_steps_per_second": 7.864, "step": 8268 }, { "epoch": 79.0, "grad_norm": 0.28201642632484436, "learning_rate": 1.05e-05, "loss": 0.0281, "step": 8374 }, { "epoch": 79.0, "eval_LOCATION_f1": 0.8378378378378377, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8157894736842105, "eval_LOCATION_recall": 0.8611111111111112, "eval_ORGANIZATION_f1": 0.7515151515151515, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6888888888888889, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.8986486486486487, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.869281045751634, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8, "eval_TIME_number": 26, "eval_TIME_precision": 0.7586206896551724, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.20406484603881836, "eval_overall_accuracy": 0.9597197898423818, "eval_overall_f1": 0.8335664335664335, "eval_overall_precision": 0.7925531914893617, "eval_overall_recall": 0.8790560471976401, "eval_runtime": 0.3821, "eval_samples_per_second": 489.378, "eval_steps_per_second": 7.851, "step": 8374 }, { "epoch": 80.0, "grad_norm": 0.2446492314338684, "learning_rate": 1e-05, "loss": 0.0275, "step": 8480 }, { "epoch": 80.0, "eval_LOCATION_f1": 0.8108108108108109, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7894736842105263, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7654320987654321, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7126436781609196, "eval_ORGANIZATION_recall": 0.8266666666666667, "eval_PERSON_f1": 0.9090909090909092, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8766233766233766, "eval_PERSON_recall": 0.9440559440559441, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7857142857142856, "eval_TIME_number": 26, "eval_TIME_precision": 0.7333333333333333, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.20445819199085236, "eval_overall_accuracy": 0.9604703527645734, "eval_overall_f1": 0.8359046283309959, "eval_overall_precision": 0.7967914438502673, "eval_overall_recall": 0.8790560471976401, "eval_runtime": 0.3791, "eval_samples_per_second": 493.229, "eval_steps_per_second": 7.913, "step": 8480 }, { "epoch": 81.0, "grad_norm": 0.8319482803344727, "learning_rate": 9.5e-06, "loss": 0.0268, "step": 8586 }, { "epoch": 81.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7435897435897435, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7160493827160493, "eval_ORGANIZATION_recall": 0.7733333333333333, "eval_PERSON_f1": 0.8963210702341137, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8589743589743589, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7857142857142856, "eval_TIME_number": 26, "eval_TIME_precision": 0.7333333333333333, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.20330098271369934, "eval_overall_accuracy": 0.9587190392794596, "eval_overall_f1": 0.8265162200282088, "eval_overall_precision": 0.7918918918918919, "eval_overall_recall": 0.8643067846607669, "eval_runtime": 0.3786, "eval_samples_per_second": 493.92, "eval_steps_per_second": 7.924, "step": 8586 }, { "epoch": 82.0, "grad_norm": 0.5323079824447632, "learning_rate": 9e-06, "loss": 0.027, "step": 8692 }, { "epoch": 82.0, "eval_LOCATION_f1": 0.8187919463087249, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7922077922077922, "eval_LOCATION_recall": 0.8472222222222222, "eval_ORGANIZATION_f1": 0.7484662576687117, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6931818181818182, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.8926174496644295, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8580645161290322, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7169811320754716, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6333333333333333, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.21506576240062714, "eval_overall_accuracy": 0.957968476357268, "eval_overall_f1": 0.8222222222222222, "eval_overall_precision": 0.7769028871391076, "eval_overall_recall": 0.8731563421828908, "eval_runtime": 0.3805, "eval_samples_per_second": 491.435, "eval_steps_per_second": 7.884, "step": 8692 }, { "epoch": 83.0, "grad_norm": 0.14162011444568634, "learning_rate": 8.500000000000002e-06, "loss": 0.0288, "step": 8798 }, { "epoch": 83.0, "eval_LOCATION_f1": 0.8108108108108109, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7894736842105263, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7421383647798742, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7023809523809523, "eval_ORGANIZATION_recall": 0.7866666666666666, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7857142857142856, "eval_TIME_number": 26, "eval_TIME_precision": 0.7333333333333333, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.2048582136631012, "eval_overall_accuracy": 0.9587190392794596, "eval_overall_f1": 0.8253521126760562, "eval_overall_precision": 0.7897574123989218, "eval_overall_recall": 0.8643067846607669, "eval_runtime": 0.378, "eval_samples_per_second": 494.671, "eval_steps_per_second": 7.936, "step": 8798 }, { "epoch": 84.0, "grad_norm": 1.106379508972168, "learning_rate": 8.000000000000001e-06, "loss": 0.0283, "step": 8904 }, { "epoch": 84.0, "eval_LOCATION_f1": 0.8219178082191781, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8108108108108109, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7577639751552795, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7093023255813954, "eval_ORGANIZATION_recall": 0.8133333333333334, "eval_PERSON_f1": 0.8993288590604027, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.864516129032258, "eval_PERSON_recall": 0.9370629370629371, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8, "eval_TIME_number": 26, "eval_TIME_precision": 0.7586206896551724, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.2061663717031479, "eval_overall_accuracy": 0.9594696022016512, "eval_overall_f1": 0.8326300984528833, "eval_overall_precision": 0.7956989247311828, "eval_overall_recall": 0.8731563421828908, "eval_runtime": 0.3799, "eval_samples_per_second": 492.292, "eval_steps_per_second": 7.898, "step": 8904 }, { "epoch": 85.0, "grad_norm": 0.4738714098930359, "learning_rate": 7.5e-06, "loss": 0.0272, "step": 9010 }, { "epoch": 85.0, "eval_LOCATION_f1": 0.8275862068965517, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.821917808219178, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.751592356687898, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7195121951219512, "eval_ORGANIZATION_recall": 0.7866666666666666, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8214285714285715, "eval_TIME_number": 26, "eval_TIME_precision": 0.7666666666666667, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.20973148941993713, "eval_overall_accuracy": 0.9587190392794596, "eval_overall_f1": 0.8328611898016997, "eval_overall_precision": 0.8010899182561307, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.3809, "eval_samples_per_second": 490.945, "eval_steps_per_second": 7.876, "step": 9010 }, { "epoch": 86.0, "grad_norm": 0.9322338104248047, "learning_rate": 7.000000000000001e-06, "loss": 0.0272, "step": 9116 }, { "epoch": 86.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7361963190184048, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6818181818181818, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7307692307692308, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6551724137931034, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7857142857142856, "eval_TIME_number": 26, "eval_TIME_precision": 0.7333333333333333, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.21231785416603088, "eval_overall_accuracy": 0.9577182887165374, "eval_overall_f1": 0.8223776223776224, "eval_overall_precision": 0.7819148936170213, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.3811, "eval_samples_per_second": 490.63, "eval_steps_per_second": 7.871, "step": 9116 }, { "epoch": 87.0, "grad_norm": 0.6208590865135193, "learning_rate": 6.5000000000000004e-06, "loss": 0.025, "step": 9222 }, { "epoch": 87.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7453416149068322, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6976744186046512, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7450980392156864, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6785714285714286, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8363636363636363, "eval_TIME_number": 26, "eval_TIME_precision": 0.7931034482758621, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.21557898819446564, "eval_overall_accuracy": 0.957968476357268, "eval_overall_f1": 0.829817158931083, "eval_overall_precision": 0.793010752688172, "eval_overall_recall": 0.8702064896755162, "eval_runtime": 0.3795, "eval_samples_per_second": 492.761, "eval_steps_per_second": 7.905, "step": 9222 }, { "epoch": 88.0, "grad_norm": 0.30410921573638916, "learning_rate": 6e-06, "loss": 0.0266, "step": 9328 }, { "epoch": 88.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7341772151898734, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6987951807228916, "eval_ORGANIZATION_recall": 0.7733333333333333, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7307692307692308, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6551724137931034, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.8214285714285715, "eval_TIME_number": 26, "eval_TIME_precision": 0.7666666666666667, "eval_TIME_recall": 0.8846153846153846, "eval_loss": 0.21103695034980774, "eval_overall_accuracy": 0.9592194145609206, "eval_overall_f1": 0.8253521126760562, "eval_overall_precision": 0.7897574123989218, "eval_overall_recall": 0.8643067846607669, "eval_runtime": 0.3802, "eval_samples_per_second": 491.818, "eval_steps_per_second": 7.89, "step": 9328 }, { "epoch": 89.0, "grad_norm": 0.8866919875144958, "learning_rate": 5.500000000000001e-06, "loss": 0.027, "step": 9434 }, { "epoch": 89.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7450980392156863, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7307692307692307, "eval_ORGANIZATION_recall": 0.76, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.7307692307692308, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6551724137931034, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7636363636363636, "eval_TIME_number": 26, "eval_TIME_precision": 0.7241379310344828, "eval_TIME_recall": 0.8076923076923077, "eval_loss": 0.2078467458486557, "eval_overall_accuracy": 0.9589692269201902, "eval_overall_f1": 0.8238636363636364, "eval_overall_precision": 0.7945205479452054, "eval_overall_recall": 0.855457227138643, "eval_runtime": 0.3827, "eval_samples_per_second": 488.686, "eval_steps_per_second": 7.84, "step": 9434 }, { "epoch": 90.0, "grad_norm": 1.0664430856704712, "learning_rate": 5e-06, "loss": 0.026, "step": 9540 }, { "epoch": 90.0, "eval_LOCATION_f1": 0.8108108108108109, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7894736842105263, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7435897435897435, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7160493827160493, "eval_ORGANIZATION_recall": 0.7733333333333333, "eval_PERSON_f1": 0.888888888888889, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8571428571428571, "eval_PERSON_recall": 0.9230769230769231, "eval_QUANTITY_f1": 0.7169811320754716, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.6333333333333333, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.20929211378097534, "eval_overall_accuracy": 0.9572179134350763, "eval_overall_f1": 0.8185654008438819, "eval_overall_precision": 0.782258064516129, "eval_overall_recall": 0.8584070796460177, "eval_runtime": 0.3786, "eval_samples_per_second": 493.926, "eval_steps_per_second": 7.924, "step": 9540 }, { "epoch": 91.0, "grad_norm": 0.24793697893619537, "learning_rate": 4.5e-06, "loss": 0.0253, "step": 9646 }, { "epoch": 91.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7468354430379747, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7108433734939759, "eval_ORGANIZATION_recall": 0.7866666666666666, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.20767036080360413, "eval_overall_accuracy": 0.9587190392794596, "eval_overall_f1": 0.8265162200282088, "eval_overall_precision": 0.7918918918918919, "eval_overall_recall": 0.8643067846607669, "eval_runtime": 0.379, "eval_samples_per_second": 493.454, "eval_steps_per_second": 7.916, "step": 9646 }, { "epoch": 92.0, "grad_norm": 0.5594867467880249, "learning_rate": 4.000000000000001e-06, "loss": 0.026, "step": 9752 }, { "epoch": 92.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7532467532467533, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7341772151898734, "eval_ORGANIZATION_recall": 0.7733333333333333, "eval_PERSON_f1": 0.8986486486486487, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.869281045751634, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.20644758641719818, "eval_overall_accuracy": 0.9597197898423818, "eval_overall_f1": 0.8295454545454547, "eval_overall_precision": 0.8, "eval_overall_recall": 0.8613569321533924, "eval_runtime": 0.3816, "eval_samples_per_second": 490.078, "eval_steps_per_second": 7.862, "step": 9752 }, { "epoch": 93.0, "grad_norm": 0.9564462304115295, "learning_rate": 3.5000000000000004e-06, "loss": 0.026, "step": 9858 }, { "epoch": 93.0, "eval_LOCATION_f1": 0.8108108108108109, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7894736842105263, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.729559748427673, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.6904761904761905, "eval_ORGANIZATION_recall": 0.7733333333333333, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.20910567045211792, "eval_overall_accuracy": 0.9584688516387291, "eval_overall_f1": 0.8213783403656821, "eval_overall_precision": 0.7849462365591398, "eval_overall_recall": 0.8613569321533924, "eval_runtime": 0.3808, "eval_samples_per_second": 491.131, "eval_steps_per_second": 7.879, "step": 9858 }, { "epoch": 94.0, "grad_norm": 0.2430673986673355, "learning_rate": 3e-06, "loss": 0.0253, "step": 9964 }, { "epoch": 94.0, "eval_LOCATION_f1": 0.8219178082191781, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8108108108108109, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7483870967741936, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.725, "eval_ORGANIZATION_recall": 0.7733333333333333, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.20926541090011597, "eval_overall_accuracy": 0.9587190392794596, "eval_overall_f1": 0.828368794326241, "eval_overall_precision": 0.7978142076502732, "eval_overall_recall": 0.8613569321533924, "eval_runtime": 0.3817, "eval_samples_per_second": 489.929, "eval_steps_per_second": 7.86, "step": 9964 }, { "epoch": 95.0, "grad_norm": 0.7441272735595703, "learning_rate": 2.5e-06, "loss": 0.0255, "step": 10070 }, { "epoch": 95.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.759493670886076, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7228915662650602, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8986486486486487, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.869281045751634, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.2069447934627533, "eval_overall_accuracy": 0.9594696022016512, "eval_overall_f1": 0.8305084745762712, "eval_overall_precision": 0.7967479674796748, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.3813, "eval_samples_per_second": 490.445, "eval_steps_per_second": 7.868, "step": 10070 }, { "epoch": 96.0, "grad_norm": 0.1459645926952362, "learning_rate": 2.0000000000000003e-06, "loss": 0.0252, "step": 10176 }, { "epoch": 96.0, "eval_LOCATION_f1": 0.8108108108108109, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.7894736842105263, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7468354430379747, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7108433734939759, "eval_ORGANIZATION_recall": 0.7866666666666666, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.20841215550899506, "eval_overall_accuracy": 0.9592194145609206, "eval_overall_f1": 0.8253521126760562, "eval_overall_precision": 0.7897574123989218, "eval_overall_recall": 0.8643067846607669, "eval_runtime": 0.3793, "eval_samples_per_second": 493.012, "eval_steps_per_second": 7.909, "step": 10176 }, { "epoch": 97.0, "grad_norm": 0.636759877204895, "learning_rate": 1.5e-06, "loss": 0.0251, "step": 10282 }, { "epoch": 97.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7483870967741936, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.725, "eval_ORGANIZATION_recall": 0.7733333333333333, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.2088312804698944, "eval_overall_accuracy": 0.9589692269201902, "eval_overall_f1": 0.8271954674220965, "eval_overall_precision": 0.7956403269754768, "eval_overall_recall": 0.8613569321533924, "eval_runtime": 0.3792, "eval_samples_per_second": 493.191, "eval_steps_per_second": 7.912, "step": 10282 }, { "epoch": 98.0, "grad_norm": 0.054069552570581436, "learning_rate": 1.0000000000000002e-06, "loss": 0.027, "step": 10388 }, { "epoch": 98.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7483870967741936, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.725, "eval_ORGANIZATION_recall": 0.7733333333333333, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.2080487757921219, "eval_overall_accuracy": 0.9589692269201902, "eval_overall_f1": 0.8271954674220965, "eval_overall_precision": 0.7956403269754768, "eval_overall_recall": 0.8613569321533924, "eval_runtime": 0.3806, "eval_samples_per_second": 491.382, "eval_steps_per_second": 7.883, "step": 10388 }, { "epoch": 99.0, "grad_norm": 0.6157656311988831, "learning_rate": 5.000000000000001e-07, "loss": 0.0239, "step": 10494 }, { "epoch": 99.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7547169811320756, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7142857142857143, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.20915448665618896, "eval_overall_accuracy": 0.9589692269201902, "eval_overall_f1": 0.828169014084507, "eval_overall_precision": 0.7924528301886793, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.386, "eval_samples_per_second": 484.484, "eval_steps_per_second": 7.772, "step": 10494 }, { "epoch": 100.0, "grad_norm": 1.120739459991455, "learning_rate": 0.0, "loss": 0.0246, "step": 10600 }, { "epoch": 100.0, "eval_LOCATION_f1": 0.816326530612245, "eval_LOCATION_number": 72, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8333333333333334, "eval_ORGANIZATION_f1": 0.7547169811320756, "eval_ORGANIZATION_number": 75, "eval_ORGANIZATION_precision": 0.7142857142857143, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8956228956228957, "eval_PERSON_number": 143, "eval_PERSON_precision": 0.8636363636363636, "eval_PERSON_recall": 0.9300699300699301, "eval_QUANTITY_f1": 0.76, "eval_QUANTITY_number": 23, "eval_QUANTITY_precision": 0.7037037037037037, "eval_QUANTITY_recall": 0.8260869565217391, "eval_TIME_f1": 0.7719298245614036, "eval_TIME_number": 26, "eval_TIME_precision": 0.7096774193548387, "eval_TIME_recall": 0.8461538461538461, "eval_loss": 0.20937226712703705, "eval_overall_accuracy": 0.9589692269201902, "eval_overall_f1": 0.828169014084507, "eval_overall_precision": 0.7924528301886793, "eval_overall_recall": 0.8672566371681416, "eval_runtime": 0.38, "eval_samples_per_second": 492.052, "eval_steps_per_second": 7.894, "step": 10600 }, { "epoch": 100.0, "step": 10600, "total_flos": 5023623771566022.0, "train_loss": 0.07358735277967633, "train_runtime": 620.9526, "train_samples_per_second": 271.679, "train_steps_per_second": 17.071 } ], "logging_steps": 500, "max_steps": 10600, "num_input_tokens_seen": 0, "num_train_epochs": 100, "save_steps": 500, "total_flos": 5023623771566022.0, "train_batch_size": 16, "trial_name": null, "trial_params": null }