{ "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": 1.6480388641357422, "learning_rate": 4.9500000000000004e-05, "loss": 0.9704, "step": 106 }, { "epoch": 1.0, "eval_LOCATION_f1": 0.0, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.0, "eval_LOCATION_recall": 0.0, "eval_ORGANIZATION_f1": 0.0, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.0, "eval_ORGANIZATION_recall": 0.0, "eval_PERSON_f1": 0.11340206185567009, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.25, "eval_PERSON_recall": 0.07333333333333333, "eval_QUANTITY_f1": 0.0, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.0, "eval_QUANTITY_recall": 0.0, "eval_TIME_f1": 0.0, "eval_TIME_number": 34, "eval_TIME_precision": 0.0, "eval_TIME_recall": 0.0, "eval_loss": 0.5637332797050476, "eval_overall_accuracy": 0.8439576667487079, "eval_overall_f1": 0.05314009661835749, "eval_overall_precision": 0.1746031746031746, "eval_overall_recall": 0.03133903133903134, "eval_runtime": 0.3779, "eval_samples_per_second": 494.871, "eval_steps_per_second": 7.939, "step": 106 }, { "epoch": 2.0, "grad_norm": 1.1430566310882568, "learning_rate": 4.9e-05, "loss": 0.4856, "step": 212 }, { "epoch": 2.0, "eval_LOCATION_f1": 0.17391304347826086, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.42105263157894735, "eval_LOCATION_recall": 0.1095890410958904, "eval_ORGANIZATION_f1": 0.09917355371900825, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.10714285714285714, "eval_ORGANIZATION_recall": 0.09230769230769231, "eval_PERSON_f1": 0.6884272997032642, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.6203208556149733, "eval_PERSON_recall": 0.7733333333333333, "eval_QUANTITY_f1": 0.07692307692307693, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.08695652173913043, "eval_QUANTITY_recall": 0.06896551724137931, "eval_TIME_f1": 0.5675675675675677, "eval_TIME_number": 34, "eval_TIME_precision": 0.525, "eval_TIME_recall": 0.6176470588235294, "eval_loss": 0.33507493138313293, "eval_overall_accuracy": 0.897612601525966, "eval_overall_f1": 0.4526627218934911, "eval_overall_precision": 0.4707692307692308, "eval_overall_recall": 0.4358974358974359, "eval_runtime": 0.374, "eval_samples_per_second": 499.997, "eval_steps_per_second": 8.021, "step": 212 }, { "epoch": 3.0, "grad_norm": 0.9914281964302063, "learning_rate": 4.85e-05, "loss": 0.2946, "step": 318 }, { "epoch": 3.0, "eval_LOCATION_f1": 0.5862068965517241, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.504950495049505, "eval_LOCATION_recall": 0.6986301369863014, "eval_ORGANIZATION_f1": 0.3787878787878788, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.373134328358209, "eval_ORGANIZATION_recall": 0.38461538461538464, "eval_PERSON_f1": 0.8322981366459627, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.7790697674418605, "eval_PERSON_recall": 0.8933333333333333, "eval_QUANTITY_f1": 0.47058823529411764, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.41025641025641024, "eval_QUANTITY_recall": 0.5517241379310345, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.20441730320453644, "eval_overall_accuracy": 0.9301009106571498, "eval_overall_f1": 0.6649282920469362, "eval_overall_precision": 0.6129807692307693, "eval_overall_recall": 0.7264957264957265, "eval_runtime": 0.3768, "eval_samples_per_second": 496.267, "eval_steps_per_second": 7.961, "step": 318 }, { "epoch": 4.0, "grad_norm": 1.5362107753753662, "learning_rate": 4.8e-05, "loss": 0.2098, "step": 424 }, { "epoch": 4.0, "eval_LOCATION_f1": 0.6987951807228916, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.6236559139784946, "eval_LOCATION_recall": 0.7945205479452054, "eval_ORGANIZATION_f1": 0.6265060240963856, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5148514851485149, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8598130841121495, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8070175438596491, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.5070422535211268, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.42857142857142855, "eval_QUANTITY_recall": 0.6206896551724138, "eval_TIME_f1": 0.8823529411764706, "eval_TIME_number": 34, "eval_TIME_precision": 0.8823529411764706, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.17477257549762726, "eval_overall_accuracy": 0.9404380999261629, "eval_overall_f1": 0.7474747474747476, "eval_overall_precision": 0.671201814058957, "eval_overall_recall": 0.8433048433048433, "eval_runtime": 0.379, "eval_samples_per_second": 493.407, "eval_steps_per_second": 7.916, "step": 424 }, { "epoch": 5.0, "grad_norm": 1.2351248264312744, "learning_rate": 4.75e-05, "loss": 0.177, "step": 530 }, { "epoch": 5.0, "eval_LOCATION_f1": 0.7672955974842768, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7093023255813954, "eval_LOCATION_recall": 0.8356164383561644, "eval_ORGANIZATION_f1": 0.6097560975609756, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5050505050505051, "eval_ORGANIZATION_recall": 0.7692307692307693, "eval_PERSON_f1": 0.8562300319488818, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8220858895705522, "eval_PERSON_recall": 0.8933333333333333, "eval_QUANTITY_f1": 0.696969696969697, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6216216216216216, "eval_QUANTITY_recall": 0.7931034482758621, "eval_TIME_f1": 0.7532467532467532, "eval_TIME_number": 34, "eval_TIME_precision": 0.6744186046511628, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.15511956810951233, "eval_overall_accuracy": 0.9492985478710313, "eval_overall_f1": 0.7625160462130937, "eval_overall_precision": 0.6939252336448598, "eval_overall_recall": 0.8461538461538461, "eval_runtime": 0.3759, "eval_samples_per_second": 497.509, "eval_steps_per_second": 7.981, "step": 530 }, { "epoch": 6.0, "grad_norm": 1.7548857927322388, "learning_rate": 4.7e-05, "loss": 0.1591, "step": 636 }, { "epoch": 6.0, "eval_LOCATION_f1": 0.7218934911242603, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.6354166666666666, "eval_LOCATION_recall": 0.8356164383561644, "eval_ORGANIZATION_f1": 0.6586826347305389, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5392156862745098, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.888178913738019, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.852760736196319, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.6, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.47058823529411764, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.6987951807228915, "eval_TIME_number": 34, "eval_TIME_precision": 0.5918367346938775, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1818351149559021, "eval_overall_accuracy": 0.9384691114939699, "eval_overall_f1": 0.7586206896551724, "eval_overall_precision": 0.6681127982646421, "eval_overall_recall": 0.8774928774928775, "eval_runtime": 0.375, "eval_samples_per_second": 498.663, "eval_steps_per_second": 8.0, "step": 636 }, { "epoch": 7.0, "grad_norm": 0.8939017057418823, "learning_rate": 4.6500000000000005e-05, "loss": 0.147, "step": 742 }, { "epoch": 7.0, "eval_LOCATION_f1": 0.7185628742514969, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.6382978723404256, "eval_LOCATION_recall": 0.821917808219178, "eval_ORGANIZATION_f1": 0.6438356164383562, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5802469135802469, "eval_ORGANIZATION_recall": 0.7230769230769231, "eval_PERSON_f1": 0.8802588996763754, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8553459119496856, "eval_PERSON_recall": 0.9066666666666666, "eval_QUANTITY_f1": 0.7272727272727273, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6486486486486487, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1428162306547165, "eval_overall_accuracy": 0.9532365247354172, "eval_overall_f1": 0.7799736495388669, "eval_overall_precision": 0.7254901960784313, "eval_overall_recall": 0.8433048433048433, "eval_runtime": 0.3769, "eval_samples_per_second": 496.191, "eval_steps_per_second": 7.96, "step": 742 }, { "epoch": 8.0, "grad_norm": 1.1140952110290527, "learning_rate": 4.600000000000001e-05, "loss": 0.138, "step": 848 }, { "epoch": 8.0, "eval_LOCATION_f1": 0.75, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.6896551724137931, "eval_LOCATION_recall": 0.821917808219178, "eval_ORGANIZATION_f1": 0.6530612244897959, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5853658536585366, "eval_ORGANIZATION_recall": 0.7384615384615385, "eval_PERSON_f1": 0.8910256410256411, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8580246913580247, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7272727272727273, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6486486486486487, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8055555555555555, "eval_TIME_number": 34, "eval_TIME_precision": 0.7631578947368421, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.13773079216480255, "eval_overall_accuracy": 0.9539748953974896, "eval_overall_f1": 0.7926023778071334, "eval_overall_precision": 0.7389162561576355, "eval_overall_recall": 0.8547008547008547, "eval_runtime": 0.3759, "eval_samples_per_second": 497.495, "eval_steps_per_second": 7.981, "step": 848 }, { "epoch": 9.0, "grad_norm": 1.3556287288665771, "learning_rate": 4.55e-05, "loss": 0.1304, "step": 954 }, { "epoch": 9.0, "eval_LOCATION_f1": 0.7741935483870966, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7317073170731707, "eval_LOCATION_recall": 0.821917808219178, "eval_ORGANIZATION_f1": 0.6405228758169935, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5568181818181818, "eval_ORGANIZATION_recall": 0.7538461538461538, "eval_PERSON_f1": 0.8867313915857605, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8616352201257862, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.716417910447761, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.631578947368421, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8333333333333333, "eval_TIME_number": 34, "eval_TIME_precision": 0.7894736842105263, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.13510727882385254, "eval_overall_accuracy": 0.9549593896135861, "eval_overall_f1": 0.7936507936507937, "eval_overall_precision": 0.7407407407407407, "eval_overall_recall": 0.8547008547008547, "eval_runtime": 0.3751, "eval_samples_per_second": 498.491, "eval_steps_per_second": 7.997, "step": 954 }, { "epoch": 10.0, "grad_norm": 0.9694278240203857, "learning_rate": 4.5e-05, "loss": 0.1262, "step": 1060 }, { "epoch": 10.0, "eval_LOCATION_f1": 0.7870967741935484, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7439024390243902, "eval_LOCATION_recall": 0.8356164383561644, "eval_ORGANIZATION_f1": 0.5864661654135338, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5735294117647058, "eval_ORGANIZATION_recall": 0.6, "eval_PERSON_f1": 0.8774193548387097, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.85, "eval_PERSON_recall": 0.9066666666666666, "eval_QUANTITY_f1": 0.696969696969697, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6216216216216216, "eval_QUANTITY_recall": 0.7931034482758621, "eval_TIME_f1": 0.8857142857142858, "eval_TIME_number": 34, "eval_TIME_precision": 0.8611111111111112, "eval_TIME_recall": 0.9117647058823529, "eval_loss": 0.13491280376911163, "eval_overall_accuracy": 0.9554516367216342, "eval_overall_f1": 0.790190735694823, "eval_overall_precision": 0.7571801566579635, "eval_overall_recall": 0.8262108262108262, "eval_runtime": 0.3755, "eval_samples_per_second": 498.046, "eval_steps_per_second": 7.99, "step": 1060 }, { "epoch": 11.0, "grad_norm": 0.7309755682945251, "learning_rate": 4.4500000000000004e-05, "loss": 0.1167, "step": 1166 }, { "epoch": 11.0, "eval_LOCATION_f1": 0.759493670886076, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7058823529411765, "eval_LOCATION_recall": 0.821917808219178, "eval_ORGANIZATION_f1": 0.6756756756756758, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6024096385542169, "eval_ORGANIZATION_recall": 0.7692307692307693, "eval_PERSON_f1": 0.8932038834951457, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8679245283018868, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.6956521739130435, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.7792207792207793, "eval_TIME_number": 34, "eval_TIME_precision": 0.6976744186046512, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.14504767954349518, "eval_overall_accuracy": 0.9549593896135861, "eval_overall_f1": 0.7936925098554534, "eval_overall_precision": 0.7365853658536585, "eval_overall_recall": 0.8603988603988604, "eval_runtime": 0.3775, "eval_samples_per_second": 495.34, "eval_steps_per_second": 7.947, "step": 1166 }, { "epoch": 12.0, "grad_norm": 0.8930480480194092, "learning_rate": 4.4000000000000006e-05, "loss": 0.1104, "step": 1272 }, { "epoch": 12.0, "eval_LOCATION_f1": 0.7239263803680981, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.6555555555555556, "eval_LOCATION_recall": 0.8082191780821918, "eval_ORGANIZATION_f1": 0.6802721088435374, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6097560975609756, "eval_ORGANIZATION_recall": 0.7692307692307693, "eval_PERSON_f1": 0.8932038834951457, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8679245283018868, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.716417910447761, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.631578947368421, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8108108108108107, "eval_TIME_number": 34, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.14461518824100494, "eval_overall_accuracy": 0.9537287718434654, "eval_overall_f1": 0.7921052631578948, "eval_overall_precision": 0.7359413202933985, "eval_overall_recall": 0.8575498575498576, "eval_runtime": 0.3751, "eval_samples_per_second": 498.517, "eval_steps_per_second": 7.998, "step": 1272 }, { "epoch": 13.0, "grad_norm": 0.6293433904647827, "learning_rate": 4.35e-05, "loss": 0.1086, "step": 1378 }, { "epoch": 13.0, "eval_LOCATION_f1": 0.751592356687898, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7023809523809523, "eval_LOCATION_recall": 0.8082191780821918, "eval_ORGANIZATION_f1": 0.6622516556291391, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5813953488372093, "eval_ORGANIZATION_recall": 0.7692307692307693, "eval_PERSON_f1": 0.9025974025974027, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.879746835443038, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7352941176470588, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6410256410256411, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8219178082191781, "eval_TIME_number": 34, "eval_TIME_precision": 0.7692307692307693, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.14269044995307922, "eval_overall_accuracy": 0.9554516367216342, "eval_overall_f1": 0.8005284015852049, "eval_overall_precision": 0.7463054187192119, "eval_overall_recall": 0.8632478632478633, "eval_runtime": 0.3775, "eval_samples_per_second": 495.368, "eval_steps_per_second": 7.947, "step": 1378 }, { "epoch": 14.0, "grad_norm": 1.0051416158676147, "learning_rate": 4.3e-05, "loss": 0.1058, "step": 1484 }, { "epoch": 14.0, "eval_LOCATION_f1": 0.7577639751552796, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.6931818181818182, "eval_LOCATION_recall": 0.8356164383561644, "eval_ORGANIZATION_f1": 0.684931506849315, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6172839506172839, "eval_ORGANIZATION_recall": 0.7692307692307693, "eval_PERSON_f1": 0.8867313915857605, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8616352201257862, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.7575757575757576, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8108108108108107, "eval_TIME_number": 34, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.13779321312904358, "eval_overall_accuracy": 0.9566822544917548, "eval_overall_f1": 0.8015873015873016, "eval_overall_precision": 0.7481481481481481, "eval_overall_recall": 0.8632478632478633, "eval_runtime": 0.376, "eval_samples_per_second": 497.317, "eval_steps_per_second": 7.978, "step": 1484 }, { "epoch": 15.0, "grad_norm": 1.7163748741149902, "learning_rate": 4.25e-05, "loss": 0.1016, "step": 1590 }, { "epoch": 15.0, "eval_LOCATION_f1": 0.7421383647798743, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.686046511627907, "eval_LOCATION_recall": 0.8082191780821918, "eval_ORGANIZATION_f1": 0.6577181208053692, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5833333333333334, "eval_ORGANIZATION_recall": 0.7538461538461538, "eval_PERSON_f1": 0.887459807073955, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8571428571428571, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7575757575757576, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8450704225352113, "eval_TIME_number": 34, "eval_TIME_precision": 0.8108108108108109, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.13695023953914642, "eval_overall_accuracy": 0.9554516367216342, "eval_overall_f1": 0.7962962962962964, "eval_overall_precision": 0.7432098765432099, "eval_overall_recall": 0.8575498575498576, "eval_runtime": 0.3774, "eval_samples_per_second": 495.557, "eval_steps_per_second": 7.95, "step": 1590 }, { "epoch": 16.0, "grad_norm": 1.4516112804412842, "learning_rate": 4.2e-05, "loss": 0.0986, "step": 1696 }, { "epoch": 16.0, "eval_LOCATION_f1": 0.7530864197530865, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.6853932584269663, "eval_LOCATION_recall": 0.8356164383561644, "eval_ORGANIZATION_f1": 0.7172413793103448, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.65, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.887459807073955, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8571428571428571, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.6493506493506493, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.5208333333333334, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7733333333333334, "eval_TIME_number": 34, "eval_TIME_precision": 0.7073170731707317, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.15388266742229462, "eval_overall_accuracy": 0.9527442776273689, "eval_overall_f1": 0.7922077922077921, "eval_overall_precision": 0.7279236276849642, "eval_overall_recall": 0.8689458689458689, "eval_runtime": 0.3786, "eval_samples_per_second": 493.9, "eval_steps_per_second": 7.924, "step": 1696 }, { "epoch": 17.0, "grad_norm": 0.6592642068862915, "learning_rate": 4.15e-05, "loss": 0.0939, "step": 1802 }, { "epoch": 17.0, "eval_LOCATION_f1": 0.7721518987341772, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7176470588235294, "eval_LOCATION_recall": 0.8356164383561644, "eval_ORGANIZATION_f1": 0.6896551724137931, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.625, "eval_ORGANIZATION_recall": 0.7692307692307693, "eval_PERSON_f1": 0.8903225806451613, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8625, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7999999999999999, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.7222222222222222, "eval_QUANTITY_recall": 0.896551724137931, "eval_TIME_f1": 0.8219178082191781, "eval_TIME_number": 34, "eval_TIME_precision": 0.7692307692307693, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.1347227245569229, "eval_overall_accuracy": 0.9584051193699237, "eval_overall_f1": 0.8122503328894807, "eval_overall_precision": 0.7625, "eval_overall_recall": 0.8689458689458689, "eval_runtime": 0.3763, "eval_samples_per_second": 496.901, "eval_steps_per_second": 7.972, "step": 1802 }, { "epoch": 18.0, "grad_norm": 1.0106799602508545, "learning_rate": 4.1e-05, "loss": 0.0915, "step": 1908 }, { "epoch": 18.0, "eval_LOCATION_f1": 0.7948717948717949, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7469879518072289, "eval_LOCATION_recall": 0.8493150684931506, "eval_ORGANIZATION_f1": 0.6756756756756758, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6024096385542169, "eval_ORGANIZATION_recall": 0.7692307692307693, "eval_PERSON_f1": 0.8932038834951457, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8679245283018868, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.746268656716418, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7945205479452054, "eval_TIME_number": 34, "eval_TIME_precision": 0.7435897435897436, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.13727760314941406, "eval_overall_accuracy": 0.9576667487078513, "eval_overall_f1": 0.8074369189907038, "eval_overall_precision": 0.7562189054726368, "eval_overall_recall": 0.8660968660968661, "eval_runtime": 0.3788, "eval_samples_per_second": 493.662, "eval_steps_per_second": 7.92, "step": 1908 }, { "epoch": 19.0, "grad_norm": 0.5167110562324524, "learning_rate": 4.05e-05, "loss": 0.0893, "step": 2014 }, { "epoch": 19.0, "eval_LOCATION_f1": 0.8051948051948052, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7654320987654321, "eval_LOCATION_recall": 0.8493150684931506, "eval_ORGANIZATION_f1": 0.6896551724137931, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.625, "eval_ORGANIZATION_recall": 0.7692307692307693, "eval_PERSON_f1": 0.8996763754045306, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8742138364779874, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.746268656716418, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.717948717948718, "eval_TIME_number": 34, "eval_TIME_precision": 0.6363636363636364, "eval_TIME_recall": 0.8235294117647058, "eval_loss": 0.13538764417171478, "eval_overall_accuracy": 0.9581589958158996, "eval_overall_f1": 0.8074369189907038, "eval_overall_precision": 0.7562189054726368, "eval_overall_recall": 0.8660968660968661, "eval_runtime": 0.3757, "eval_samples_per_second": 497.713, "eval_steps_per_second": 7.985, "step": 2014 }, { "epoch": 20.0, "grad_norm": 0.4847421646118164, "learning_rate": 4e-05, "loss": 0.0836, "step": 2120 }, { "epoch": 20.0, "eval_LOCATION_f1": 0.7692307692307693, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7228915662650602, "eval_LOCATION_recall": 0.821917808219178, "eval_ORGANIZATION_f1": 0.7074829931972789, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6341463414634146, "eval_ORGANIZATION_recall": 0.8, "eval_PERSON_f1": 0.8838709677419355, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.85625, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.7575757575757576, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8333333333333333, "eval_TIME_number": 34, "eval_TIME_precision": 0.7894736842105263, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.14133651554584503, "eval_overall_accuracy": 0.9581589958158996, "eval_overall_f1": 0.8095872170439414, "eval_overall_precision": 0.76, "eval_overall_recall": 0.8660968660968661, "eval_runtime": 0.379, "eval_samples_per_second": 493.465, "eval_steps_per_second": 7.917, "step": 2120 }, { "epoch": 21.0, "grad_norm": 1.5779582262039185, "learning_rate": 3.9500000000000005e-05, "loss": 0.08, "step": 2226 }, { "epoch": 21.0, "eval_LOCATION_f1": 0.7643312101910829, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.821917808219178, "eval_ORGANIZATION_f1": 0.684931506849315, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6172839506172839, "eval_ORGANIZATION_recall": 0.7692307692307693, "eval_PERSON_f1": 0.9055374592833875, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8853503184713376, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7272727272727273, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6486486486486487, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8405797101449276, "eval_TIME_number": 34, "eval_TIME_precision": 0.8285714285714286, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1396959125995636, "eval_overall_accuracy": 0.9584051193699237, "eval_overall_f1": 0.810738255033557, "eval_overall_precision": 0.766497461928934, "eval_overall_recall": 0.8603988603988604, "eval_runtime": 0.3754, "eval_samples_per_second": 498.078, "eval_steps_per_second": 7.991, "step": 2226 }, { "epoch": 22.0, "grad_norm": 1.488678216934204, "learning_rate": 3.9000000000000006e-05, "loss": 0.0805, "step": 2332 }, { "epoch": 22.0, "eval_LOCATION_f1": 0.7973856209150326, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7625, "eval_LOCATION_recall": 0.8356164383561644, "eval_ORGANIZATION_f1": 0.6805555555555556, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.620253164556962, "eval_ORGANIZATION_recall": 0.7538461538461538, "eval_PERSON_f1": 0.9084967320261438, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8910256410256411, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7575757575757576, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8333333333333333, "eval_TIME_number": 34, "eval_TIME_precision": 0.7894736842105263, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.13735245168209076, "eval_overall_accuracy": 0.9606202313561408, "eval_overall_f1": 0.8205128205128205, "eval_overall_precision": 0.7794871794871795, "eval_overall_recall": 0.8660968660968661, "eval_runtime": 0.3761, "eval_samples_per_second": 497.224, "eval_steps_per_second": 7.977, "step": 2332 }, { "epoch": 23.0, "grad_norm": 0.8602825999259949, "learning_rate": 3.85e-05, "loss": 0.0802, "step": 2438 }, { "epoch": 23.0, "eval_LOCATION_f1": 0.7870967741935484, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7439024390243902, "eval_LOCATION_recall": 0.8356164383561644, "eval_ORGANIZATION_f1": 0.7058823529411765, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6136363636363636, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.8903225806451613, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8625, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7936507936507937, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.7352941176470589, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8985507246376812, "eval_TIME_number": 34, "eval_TIME_precision": 0.8857142857142857, "eval_TIME_recall": 0.9117647058823529, "eval_loss": 0.1325836330652237, "eval_overall_accuracy": 0.9608663549101649, "eval_overall_f1": 0.824, "eval_overall_precision": 0.7744360902255639, "eval_overall_recall": 0.8803418803418803, "eval_runtime": 0.3799, "eval_samples_per_second": 492.244, "eval_steps_per_second": 7.897, "step": 2438 }, { "epoch": 24.0, "grad_norm": 0.8622602820396423, "learning_rate": 3.8e-05, "loss": 0.0758, "step": 2544 }, { "epoch": 24.0, "eval_LOCATION_f1": 0.759493670886076, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7058823529411765, "eval_LOCATION_recall": 0.821917808219178, "eval_ORGANIZATION_f1": 0.7058823529411765, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6136363636363636, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.9144736842105262, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.9025974025974026, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.746268656716418, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7894736842105262, "eval_TIME_number": 34, "eval_TIME_precision": 0.7142857142857143, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.14558185636997223, "eval_overall_accuracy": 0.9571745015998031, "eval_overall_f1": 0.8126649076517152, "eval_overall_precision": 0.7567567567567568, "eval_overall_recall": 0.8774928774928775, "eval_runtime": 0.3752, "eval_samples_per_second": 498.36, "eval_steps_per_second": 7.995, "step": 2544 }, { "epoch": 25.0, "grad_norm": 1.5843734741210938, "learning_rate": 3.7500000000000003e-05, "loss": 0.0702, "step": 2650 }, { "epoch": 25.0, "eval_LOCATION_f1": 0.7870967741935484, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7439024390243902, "eval_LOCATION_recall": 0.8356164383561644, "eval_ORGANIZATION_f1": 0.6878980891719746, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5869565217391305, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.896103896103896, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8734177215189873, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.6857142857142856, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.5853658536585366, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8571428571428571, "eval_TIME_number": 34, "eval_TIME_precision": 0.8333333333333334, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.15064221620559692, "eval_overall_accuracy": 0.9564361309377307, "eval_overall_f1": 0.8078947368421052, "eval_overall_precision": 0.7506112469437652, "eval_overall_recall": 0.8746438746438746, "eval_runtime": 0.3763, "eval_samples_per_second": 496.974, "eval_steps_per_second": 7.973, "step": 2650 }, { "epoch": 26.0, "grad_norm": 0.8419085741043091, "learning_rate": 3.7e-05, "loss": 0.0718, "step": 2756 }, { "epoch": 26.0, "eval_LOCATION_f1": 0.8104575163398693, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.775, "eval_LOCATION_recall": 0.8493150684931506, "eval_ORGANIZATION_f1": 0.7034482758620688, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6375, "eval_ORGANIZATION_recall": 0.7846153846153846, "eval_PERSON_f1": 0.9049180327868852, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8903225806451613, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.676056338028169, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.5714285714285714, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8219178082191781, "eval_TIME_number": 34, "eval_TIME_precision": 0.7692307692307693, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.1452375054359436, "eval_overall_accuracy": 0.9591434900319961, "eval_overall_f1": 0.8165997322623829, "eval_overall_precision": 0.7702020202020202, "eval_overall_recall": 0.8689458689458689, "eval_runtime": 0.3768, "eval_samples_per_second": 496.331, "eval_steps_per_second": 7.963, "step": 2756 }, { "epoch": 27.0, "grad_norm": 0.5116803646087646, "learning_rate": 3.65e-05, "loss": 0.0724, "step": 2862 }, { "epoch": 27.0, "eval_LOCATION_f1": 0.7820512820512822, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7349397590361446, "eval_LOCATION_recall": 0.8356164383561644, "eval_ORGANIZATION_f1": 0.7248322147651007, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6428571428571429, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.9049180327868852, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8903225806451613, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7042253521126761, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.5952380952380952, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7631578947368423, "eval_TIME_number": 34, "eval_TIME_precision": 0.6904761904761905, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.15233390033245087, "eval_overall_accuracy": 0.9556977602756583, "eval_overall_f1": 0.8110964332892999, "eval_overall_precision": 0.7561576354679803, "eval_overall_recall": 0.8746438746438746, "eval_runtime": 0.3761, "eval_samples_per_second": 497.196, "eval_steps_per_second": 7.976, "step": 2862 }, { "epoch": 28.0, "grad_norm": 1.6091852188110352, "learning_rate": 3.6e-05, "loss": 0.0662, "step": 2968 }, { "epoch": 28.0, "eval_LOCATION_f1": 0.8344370860927152, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8076923076923077, "eval_LOCATION_recall": 0.863013698630137, "eval_ORGANIZATION_f1": 0.7297297297297298, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6506024096385542, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.896103896103896, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8734177215189873, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7575757575757576, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7631578947368423, "eval_TIME_number": 34, "eval_TIME_precision": 0.6904761904761905, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.14936165511608124, "eval_overall_accuracy": 0.9603741078021166, "eval_overall_f1": 0.8251001335113485, "eval_overall_precision": 0.7763819095477387, "eval_overall_recall": 0.8803418803418803, "eval_runtime": 0.3772, "eval_samples_per_second": 495.8, "eval_steps_per_second": 7.954, "step": 2968 }, { "epoch": 29.0, "grad_norm": 0.4799008071422577, "learning_rate": 3.55e-05, "loss": 0.0661, "step": 3074 }, { "epoch": 29.0, "eval_LOCATION_f1": 0.8211920529801324, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7948717948717948, "eval_LOCATION_recall": 0.8493150684931506, "eval_ORGANIZATION_f1": 0.7417218543046358, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6511627906976745, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9055374592833875, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8853503184713376, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7619047619047619, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.7058823529411765, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8219178082191781, "eval_TIME_number": 34, "eval_TIME_precision": 0.7692307692307693, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.1376074105501175, "eval_overall_accuracy": 0.9633275904504061, "eval_overall_f1": 0.8348993288590604, "eval_overall_precision": 0.7893401015228426, "eval_overall_recall": 0.886039886039886, "eval_runtime": 0.3767, "eval_samples_per_second": 496.403, "eval_steps_per_second": 7.964, "step": 3074 }, { "epoch": 30.0, "grad_norm": 3.4121408462524414, "learning_rate": 3.5e-05, "loss": 0.0671, "step": 3180 }, { "epoch": 30.0, "eval_LOCATION_f1": 0.847682119205298, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8205128205128205, "eval_LOCATION_recall": 0.8767123287671232, "eval_ORGANIZATION_f1": 0.7111111111111111, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6857142857142857, "eval_ORGANIZATION_recall": 0.7384615384615385, "eval_PERSON_f1": 0.896774193548387, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.86875, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7936507936507937, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.7352941176470589, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8333333333333333, "eval_TIME_number": 34, "eval_TIME_precision": 0.7894736842105263, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.1364855319261551, "eval_overall_accuracy": 0.9616047255722373, "eval_overall_f1": 0.8372093023255812, "eval_overall_precision": 0.8052631578947368, "eval_overall_recall": 0.8717948717948718, "eval_runtime": 0.377, "eval_samples_per_second": 496.081, "eval_steps_per_second": 7.959, "step": 3180 }, { "epoch": 31.0, "grad_norm": 0.4697336256504059, "learning_rate": 3.45e-05, "loss": 0.0627, "step": 3286 }, { "epoch": 31.0, "eval_LOCATION_f1": 0.8684210526315789, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8354430379746836, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.7808219178082192, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.7037037037037037, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8990228013029316, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8789808917197452, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7575757575757576, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8333333333333333, "eval_TIME_number": 34, "eval_TIME_precision": 0.7894736842105263, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.1380578875541687, "eval_overall_accuracy": 0.9655427024366232, "eval_overall_f1": 0.8506056527590848, "eval_overall_precision": 0.8061224489795918, "eval_overall_recall": 0.9002849002849003, "eval_runtime": 0.3763, "eval_samples_per_second": 497.01, "eval_steps_per_second": 7.973, "step": 3286 }, { "epoch": 32.0, "grad_norm": 0.09275401383638382, "learning_rate": 3.4000000000000007e-05, "loss": 0.0599, "step": 3392 }, { "epoch": 32.0, "eval_LOCATION_f1": 0.8051948051948052, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7654320987654321, "eval_LOCATION_recall": 0.8493150684931506, "eval_ORGANIZATION_f1": 0.6993865030674845, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5816326530612245, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8737864077669903, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8490566037735849, "eval_PERSON_recall": 0.9, "eval_QUANTITY_f1": 0.7812500000000001, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.7142857142857143, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8, "eval_TIME_number": 34, "eval_TIME_precision": 0.7317073170731707, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.15592296421527863, "eval_overall_accuracy": 0.9559438838296825, "eval_overall_f1": 0.807843137254902, "eval_overall_precision": 0.7463768115942029, "eval_overall_recall": 0.8803418803418803, "eval_runtime": 0.3756, "eval_samples_per_second": 497.918, "eval_steps_per_second": 7.988, "step": 3392 }, { "epoch": 33.0, "grad_norm": 0.8940030932426453, "learning_rate": 3.35e-05, "loss": 0.0639, "step": 3498 }, { "epoch": 33.0, "eval_LOCATION_f1": 0.8400000000000001, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8181818181818182, "eval_LOCATION_recall": 0.863013698630137, "eval_ORGANIZATION_f1": 0.6838709677419356, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.5888888888888889, "eval_ORGANIZATION_recall": 0.8153846153846154, "eval_PERSON_f1": 0.8726114649681529, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8353658536585366, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.7692307692307692, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6944444444444444, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7837837837837837, "eval_TIME_number": 34, "eval_TIME_precision": 0.725, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.15322409570217133, "eval_overall_accuracy": 0.9576667487078513, "eval_overall_f1": 0.8100263852242744, "eval_overall_precision": 0.7542997542997543, "eval_overall_recall": 0.8746438746438746, "eval_runtime": 0.3774, "eval_samples_per_second": 495.453, "eval_steps_per_second": 7.948, "step": 3498 }, { "epoch": 34.0, "grad_norm": 0.9726054072380066, "learning_rate": 3.3e-05, "loss": 0.0611, "step": 3604 }, { "epoch": 34.0, "eval_LOCATION_f1": 0.847682119205298, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8205128205128205, "eval_LOCATION_recall": 0.8767123287671232, "eval_ORGANIZATION_f1": 0.7333333333333334, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6470588235294118, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8846153846153846, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8518518518518519, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7575757575757576, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8333333333333333, "eval_TIME_number": 34, "eval_TIME_precision": 0.7894736842105263, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.1448259800672531, "eval_overall_accuracy": 0.9616047255722373, "eval_overall_f1": 0.8308921438082557, "eval_overall_precision": 0.78, "eval_overall_recall": 0.8888888888888888, "eval_runtime": 0.3772, "eval_samples_per_second": 495.718, "eval_steps_per_second": 7.953, "step": 3604 }, { "epoch": 35.0, "grad_norm": 0.49669384956359863, "learning_rate": 3.2500000000000004e-05, "loss": 0.0601, "step": 3710 }, { "epoch": 35.0, "eval_LOCATION_f1": 0.8289473684210525, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7974683544303798, "eval_LOCATION_recall": 0.863013698630137, "eval_ORGANIZATION_f1": 0.7225806451612903, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6222222222222222, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8932038834951457, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8679245283018868, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.625, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7837837837837837, "eval_TIME_number": 34, "eval_TIME_precision": 0.725, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.15617811679840088, "eval_overall_accuracy": 0.9559438838296825, "eval_overall_f1": 0.8194993412384717, "eval_overall_precision": 0.7622549019607843, "eval_overall_recall": 0.886039886039886, "eval_runtime": 0.3794, "eval_samples_per_second": 492.904, "eval_steps_per_second": 7.908, "step": 3710 }, { "epoch": 36.0, "grad_norm": 1.0704704523086548, "learning_rate": 3.2000000000000005e-05, "loss": 0.061, "step": 3816 }, { "epoch": 36.0, "eval_LOCATION_f1": 0.88, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8571428571428571, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.76, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6705882352941176, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8867313915857605, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8616352201257862, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.746268656716418, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7466666666666667, "eval_TIME_number": 34, "eval_TIME_precision": 0.6829268292682927, "eval_TIME_recall": 0.8235294117647058, "eval_loss": 0.15318439900875092, "eval_overall_accuracy": 0.958897366477972, "eval_overall_f1": 0.833555259653795, "eval_overall_precision": 0.7825, "eval_overall_recall": 0.8917378917378918, "eval_runtime": 0.3758, "eval_samples_per_second": 497.576, "eval_steps_per_second": 7.983, "step": 3816 }, { "epoch": 37.0, "grad_norm": 0.6337342262268066, "learning_rate": 3.15e-05, "loss": 0.0575, "step": 3922 }, { "epoch": 37.0, "eval_LOCATION_f1": 0.7999999999999999, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7560975609756098, "eval_LOCATION_recall": 0.8493150684931506, "eval_ORGANIZATION_f1": 0.7080745341614907, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.59375, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8932038834951457, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8679245283018868, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.746268656716418, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7733333333333334, "eval_TIME_number": 34, "eval_TIME_precision": 0.7073170731707317, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.15566201508045197, "eval_overall_accuracy": 0.9561900073837066, "eval_overall_f1": 0.8109517601043025, "eval_overall_precision": 0.7475961538461539, "eval_overall_recall": 0.886039886039886, "eval_runtime": 0.3755, "eval_samples_per_second": 497.974, "eval_steps_per_second": 7.989, "step": 3922 }, { "epoch": 38.0, "grad_norm": 0.656589925289154, "learning_rate": 3.1e-05, "loss": 0.0548, "step": 4028 }, { "epoch": 38.0, "eval_LOCATION_f1": 0.855263157894737, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8227848101265823, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7092198581560285, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6578947368421053, "eval_ORGANIZATION_recall": 0.7692307692307693, "eval_PERSON_f1": 0.9025974025974027, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.879746835443038, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.6857142857142856, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.5853658536585366, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.7733333333333334, "eval_TIME_number": 34, "eval_TIME_precision": 0.7073170731707317, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.15720093250274658, "eval_overall_accuracy": 0.9576667487078513, "eval_overall_f1": 0.8230563002680965, "eval_overall_precision": 0.7772151898734178, "eval_overall_recall": 0.8746438746438746, "eval_runtime": 0.3786, "eval_samples_per_second": 493.924, "eval_steps_per_second": 7.924, "step": 4028 }, { "epoch": 39.0, "grad_norm": 0.499746173620224, "learning_rate": 3.05e-05, "loss": 0.0563, "step": 4134 }, { "epoch": 39.0, "eval_LOCATION_f1": 0.847682119205298, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8205128205128205, "eval_LOCATION_recall": 0.8767123287671232, "eval_ORGANIZATION_f1": 0.7402597402597403, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6404494382022472, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.9055374592833875, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8853503184713376, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7619047619047619, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.7058823529411765, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.7945205479452054, "eval_TIME_number": 34, "eval_TIME_precision": 0.7435897435897436, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1509544402360916, "eval_overall_accuracy": 0.9608663549101649, "eval_overall_f1": 0.8368983957219251, "eval_overall_precision": 0.7884130982367759, "eval_overall_recall": 0.8917378917378918, "eval_runtime": 0.3757, "eval_samples_per_second": 497.719, "eval_steps_per_second": 7.985, "step": 4134 }, { "epoch": 40.0, "grad_norm": 0.6194659471511841, "learning_rate": 3e-05, "loss": 0.0552, "step": 4240 }, { "epoch": 40.0, "eval_LOCATION_f1": 0.8366013071895425, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8, "eval_LOCATION_recall": 0.8767123287671232, "eval_ORGANIZATION_f1": 0.7225806451612903, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6222222222222222, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.896103896103896, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8734177215189873, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7936507936507937, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.7352941176470589, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7837837837837837, "eval_TIME_number": 34, "eval_TIME_precision": 0.725, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.14581945538520813, "eval_overall_accuracy": 0.9606202313561408, "eval_overall_f1": 0.8286852589641434, "eval_overall_precision": 0.7761194029850746, "eval_overall_recall": 0.8888888888888888, "eval_runtime": 0.3777, "eval_samples_per_second": 495.091, "eval_steps_per_second": 7.943, "step": 4240 }, { "epoch": 41.0, "grad_norm": 0.409952849149704, "learning_rate": 2.95e-05, "loss": 0.0551, "step": 4346 }, { "epoch": 41.0, "eval_LOCATION_f1": 0.8441558441558442, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8024691358024691, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7189542483660131, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.625, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8903225806451613, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8625, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7692307692307692, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6944444444444444, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8055555555555555, "eval_TIME_number": 34, "eval_TIME_precision": 0.7631578947368421, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.14438895881175995, "eval_overall_accuracy": 0.9616047255722373, "eval_overall_f1": 0.8275862068965517, "eval_overall_precision": 0.7741935483870968, "eval_overall_recall": 0.8888888888888888, "eval_runtime": 0.3738, "eval_samples_per_second": 500.23, "eval_steps_per_second": 8.025, "step": 4346 }, { "epoch": 42.0, "grad_norm": 0.3523612320423126, "learning_rate": 2.9e-05, "loss": 0.0518, "step": 4452 }, { "epoch": 42.0, "eval_LOCATION_f1": 0.8289473684210525, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.7974683544303798, "eval_LOCATION_recall": 0.863013698630137, "eval_ORGANIZATION_f1": 0.7581699346405228, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6590909090909091, "eval_ORGANIZATION_recall": 0.8923076923076924, "eval_PERSON_f1": 0.8932038834951457, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8679245283018868, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.6857142857142856, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.5853658536585366, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8219178082191781, "eval_TIME_number": 34, "eval_TIME_precision": 0.7692307692307693, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.16775819659233093, "eval_overall_accuracy": 0.9584051193699237, "eval_overall_f1": 0.8269484808454426, "eval_overall_precision": 0.770935960591133, "eval_overall_recall": 0.8917378917378918, "eval_runtime": 0.3789, "eval_samples_per_second": 493.546, "eval_steps_per_second": 7.918, "step": 4452 }, { "epoch": 43.0, "grad_norm": 0.7837867140769958, "learning_rate": 2.8499999999999998e-05, "loss": 0.0493, "step": 4558 }, { "epoch": 43.0, "eval_LOCATION_f1": 0.8344370860927152, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8076923076923077, "eval_LOCATION_recall": 0.863013698630137, "eval_ORGANIZATION_f1": 0.7567567567567568, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6746987951807228, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8990228013029316, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8789808917197452, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.6666666666666666, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.575, "eval_QUANTITY_recall": 0.7931034482758621, "eval_TIME_f1": 0.8493150684931507, "eval_TIME_number": 34, "eval_TIME_precision": 0.7948717948717948, "eval_TIME_recall": 0.9117647058823529, "eval_loss": 0.1582229733467102, "eval_overall_accuracy": 0.9618508491262614, "eval_overall_f1": 0.8315508021390374, "eval_overall_precision": 0.783375314861461, "eval_overall_recall": 0.886039886039886, "eval_runtime": 0.3754, "eval_samples_per_second": 498.081, "eval_steps_per_second": 7.991, "step": 4558 }, { "epoch": 44.0, "grad_norm": 0.7167583703994751, "learning_rate": 2.8000000000000003e-05, "loss": 0.0499, "step": 4664 }, { "epoch": 44.0, "eval_LOCATION_f1": 0.8493150684931505, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8493150684931506, "eval_LOCATION_recall": 0.8493150684931506, "eval_ORGANIZATION_f1": 0.7702702702702702, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6867469879518072, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.9114754098360656, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.896774193548387, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7619047619047619, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.7058823529411765, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8405797101449276, "eval_TIME_number": 34, "eval_TIME_precision": 0.8285714285714286, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.14848025143146515, "eval_overall_accuracy": 0.9640659611124784, "eval_overall_f1": 0.8508891928864568, "eval_overall_precision": 0.8184210526315789, "eval_overall_recall": 0.886039886039886, "eval_runtime": 0.3761, "eval_samples_per_second": 497.195, "eval_steps_per_second": 7.976, "step": 4664 }, { "epoch": 45.0, "grad_norm": 0.9075788259506226, "learning_rate": 2.7500000000000004e-05, "loss": 0.0486, "step": 4770 }, { "epoch": 45.0, "eval_LOCATION_f1": 0.8456375838926175, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8289473684210527, "eval_LOCATION_recall": 0.863013698630137, "eval_ORGANIZATION_f1": 0.75, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6551724137931034, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.9084967320261438, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8910256410256411, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.6956521739130435, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.7945205479452054, "eval_TIME_number": 34, "eval_TIME_precision": 0.7435897435897436, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.15773575007915497, "eval_overall_accuracy": 0.9608663549101649, "eval_overall_f1": 0.8331108144192256, "eval_overall_precision": 0.7839195979899497, "eval_overall_recall": 0.8888888888888888, "eval_runtime": 0.3751, "eval_samples_per_second": 498.585, "eval_steps_per_second": 7.999, "step": 4770 }, { "epoch": 46.0, "grad_norm": 0.3946699798107147, "learning_rate": 2.7000000000000002e-05, "loss": 0.0449, "step": 4876 }, { "epoch": 46.0, "eval_LOCATION_f1": 0.855263157894737, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8227848101265823, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7432432432432433, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6626506024096386, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9013157894736842, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8896103896103896, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.7352941176470588, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6410256410256411, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8450704225352113, "eval_TIME_number": 34, "eval_TIME_precision": 0.8108108108108109, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.15104267001152039, "eval_overall_accuracy": 0.9633275904504061, "eval_overall_f1": 0.8398384925975775, "eval_overall_precision": 0.7959183673469388, "eval_overall_recall": 0.8888888888888888, "eval_runtime": 0.3745, "eval_samples_per_second": 499.33, "eval_steps_per_second": 8.011, "step": 4876 }, { "epoch": 47.0, "grad_norm": 0.5078842043876648, "learning_rate": 2.6500000000000004e-05, "loss": 0.0473, "step": 4982 }, { "epoch": 47.0, "eval_LOCATION_f1": 0.8496732026143791, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8125, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7354838709677419, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6333333333333333, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.9025974025974027, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.879746835443038, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7692307692307692, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6944444444444444, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7631578947368423, "eval_TIME_number": 34, "eval_TIME_precision": 0.6904761904761905, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1570935845375061, "eval_overall_accuracy": 0.9618508491262614, "eval_overall_f1": 0.8322324966974901, "eval_overall_precision": 0.7758620689655172, "eval_overall_recall": 0.8974358974358975, "eval_runtime": 0.3787, "eval_samples_per_second": 493.833, "eval_steps_per_second": 7.922, "step": 4982 }, { "epoch": 48.0, "grad_norm": 3.7828359603881836, "learning_rate": 2.6000000000000002e-05, "loss": 0.0491, "step": 5088 }, { "epoch": 48.0, "eval_LOCATION_f1": 0.8496732026143791, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8125, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7402597402597403, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6404494382022472, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8903225806451613, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8625, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.6849315068493151, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.5681818181818182, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7894736842105262, "eval_TIME_number": 34, "eval_TIME_precision": 0.7142857142857143, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.1721905767917633, "eval_overall_accuracy": 0.9571745015998031, "eval_overall_f1": 0.8224543080939948, "eval_overall_precision": 0.7590361445783133, "eval_overall_recall": 0.8974358974358975, "eval_runtime": 0.3762, "eval_samples_per_second": 497.05, "eval_steps_per_second": 7.974, "step": 5088 }, { "epoch": 49.0, "grad_norm": 0.6315198540687561, "learning_rate": 2.5500000000000003e-05, "loss": 0.0471, "step": 5194 }, { "epoch": 49.0, "eval_LOCATION_f1": 0.8609271523178808, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8333333333333334, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7682119205298013, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6744186046511628, "eval_ORGANIZATION_recall": 0.8923076923076924, "eval_PERSON_f1": 0.9055374592833875, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8853503184713376, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7352941176470588, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6410256410256411, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8108108108108107, "eval_TIME_number": 34, "eval_TIME_precision": 0.75, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.16492243111133575, "eval_overall_accuracy": 0.9623430962343096, "eval_overall_f1": 0.844207723035952, "eval_overall_precision": 0.7925, "eval_overall_recall": 0.9031339031339032, "eval_runtime": 0.375, "eval_samples_per_second": 498.618, "eval_steps_per_second": 7.999, "step": 5194 }, { "epoch": 50.0, "grad_norm": 2.063199996948242, "learning_rate": 2.5e-05, "loss": 0.0454, "step": 5300 }, { "epoch": 50.0, "eval_LOCATION_f1": 0.8421052631578948, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.810126582278481, "eval_LOCATION_recall": 0.8767123287671232, "eval_ORGANIZATION_f1": 0.7755102040816327, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6951219512195121, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8996763754045306, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8742138364779874, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7812500000000001, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.7142857142857143, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7945205479452054, "eval_TIME_number": 34, "eval_TIME_precision": 0.7435897435897436, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.15503931045532227, "eval_overall_accuracy": 0.9645582082205267, "eval_overall_f1": 0.8429530201342281, "eval_overall_precision": 0.7969543147208121, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3783, "eval_samples_per_second": 494.287, "eval_steps_per_second": 7.93, "step": 5300 }, { "epoch": 51.0, "grad_norm": 0.47312086820602417, "learning_rate": 2.45e-05, "loss": 0.0453, "step": 5406 }, { "epoch": 51.0, "eval_LOCATION_f1": 0.8421052631578948, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.810126582278481, "eval_LOCATION_recall": 0.8767123287671232, "eval_ORGANIZATION_f1": 0.7659574468085107, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.7105263157894737, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.9025974025974027, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.879746835443038, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.716417910447761, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.631578947368421, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.7945205479452054, "eval_TIME_number": 34, "eval_TIME_precision": 0.7435897435897436, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.15570588409900665, "eval_overall_accuracy": 0.9638198375584544, "eval_overall_f1": 0.8367071524966261, "eval_overall_precision": 0.7948717948717948, "eval_overall_recall": 0.8831908831908832, "eval_runtime": 0.3773, "eval_samples_per_second": 495.687, "eval_steps_per_second": 7.952, "step": 5406 }, { "epoch": 52.0, "grad_norm": 0.6043301224708557, "learning_rate": 2.4e-05, "loss": 0.0467, "step": 5512 }, { "epoch": 52.0, "eval_LOCATION_f1": 0.8590604026845637, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8421052631578947, "eval_LOCATION_recall": 0.8767123287671232, "eval_ORGANIZATION_f1": 0.7581699346405228, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6590909090909091, "eval_ORGANIZATION_recall": 0.8923076923076924, "eval_PERSON_f1": 0.9114754098360656, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.896774193548387, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.676056338028169, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.5714285714285714, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.7945205479452054, "eval_TIME_number": 34, "eval_TIME_precision": 0.7435897435897436, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.16677530109882355, "eval_overall_accuracy": 0.9606202313561408, "eval_overall_f1": 0.8362183754993342, "eval_overall_precision": 0.785, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3783, "eval_samples_per_second": 494.283, "eval_steps_per_second": 7.93, "step": 5512 }, { "epoch": 53.0, "grad_norm": 0.5334353446960449, "learning_rate": 2.35e-05, "loss": 0.0415, "step": 5618 }, { "epoch": 53.0, "eval_LOCATION_f1": 0.847682119205298, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8205128205128205, "eval_LOCATION_recall": 0.8767123287671232, "eval_ORGANIZATION_f1": 0.7450980392156863, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6477272727272727, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.9090909090909091, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8860759493670886, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.676056338028169, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.5714285714285714, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.7733333333333334, "eval_TIME_number": 34, "eval_TIME_precision": 0.7073170731707317, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.16488516330718994, "eval_overall_accuracy": 0.9598818606940684, "eval_overall_f1": 0.828496042216359, "eval_overall_precision": 0.7714987714987716, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3783, "eval_samples_per_second": 494.268, "eval_steps_per_second": 7.929, "step": 5618 }, { "epoch": 54.0, "grad_norm": 0.7248595356941223, "learning_rate": 2.3000000000000003e-05, "loss": 0.0428, "step": 5724 }, { "epoch": 54.0, "eval_LOCATION_f1": 0.8724832214765101, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8552631578947368, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7532467532467532, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.651685393258427, "eval_ORGANIZATION_recall": 0.8923076923076924, "eval_PERSON_f1": 0.9155844155844156, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8924050632911392, "eval_PERSON_recall": 0.94, "eval_QUANTITY_f1": 0.7058823529411765, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6153846153846154, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1656116396188736, "eval_overall_accuracy": 0.9618508491262614, "eval_overall_f1": 0.8453333333333334, "eval_overall_precision": 0.7944862155388471, "eval_overall_recall": 0.9031339031339032, "eval_runtime": 0.3763, "eval_samples_per_second": 496.973, "eval_steps_per_second": 7.973, "step": 5724 }, { "epoch": 55.0, "grad_norm": 0.7314643859863281, "learning_rate": 2.25e-05, "loss": 0.042, "step": 5830 }, { "epoch": 55.0, "eval_LOCATION_f1": 0.8627450980392157, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.825, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.7581699346405228, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6590909090909091, "eval_ORGANIZATION_recall": 0.8923076923076924, "eval_PERSON_f1": 0.9055374592833875, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8853503184713376, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.625, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8285714285714286, "eval_TIME_number": 34, "eval_TIME_precision": 0.8055555555555556, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.16774289309978485, "eval_overall_accuracy": 0.9603741078021166, "eval_overall_f1": 0.8430851063829787, "eval_overall_precision": 0.7905236907730673, "eval_overall_recall": 0.9031339031339032, "eval_runtime": 0.377, "eval_samples_per_second": 496.083, "eval_steps_per_second": 7.959, "step": 5830 }, { "epoch": 56.0, "grad_norm": 1.3385694026947021, "learning_rate": 2.2000000000000003e-05, "loss": 0.0412, "step": 5936 }, { "epoch": 56.0, "eval_LOCATION_f1": 0.8724832214765101, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8552631578947368, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7862068965517242, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.7125, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8925081433224756, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8726114649681529, "eval_PERSON_recall": 0.9133333333333333, "eval_QUANTITY_f1": 0.7384615384615385, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6666666666666666, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8285714285714286, "eval_TIME_number": 34, "eval_TIME_precision": 0.8055555555555556, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.15095703303813934, "eval_overall_accuracy": 0.9655427024366232, "eval_overall_f1": 0.8478260869565218, "eval_overall_precision": 0.8103896103896104, "eval_overall_recall": 0.8888888888888888, "eval_runtime": 0.3771, "eval_samples_per_second": 495.91, "eval_steps_per_second": 7.956, "step": 5936 }, { "epoch": 57.0, "grad_norm": 1.0661405324935913, "learning_rate": 2.15e-05, "loss": 0.0409, "step": 6042 }, { "epoch": 57.0, "eval_LOCATION_f1": 0.8590604026845637, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8421052631578947, "eval_LOCATION_recall": 0.8767123287671232, "eval_ORGANIZATION_f1": 0.7552447552447553, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6923076923076923, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.9019607843137256, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8846153846153846, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.787878787878788, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.7027027027027027, "eval_QUANTITY_recall": 0.896551724137931, "eval_TIME_f1": 0.8571428571428571, "eval_TIME_number": 34, "eval_TIME_precision": 0.8333333333333334, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.15901347994804382, "eval_overall_accuracy": 0.965296578882599, "eval_overall_f1": 0.8501362397820162, "eval_overall_precision": 0.814621409921671, "eval_overall_recall": 0.8888888888888888, "eval_runtime": 0.3774, "eval_samples_per_second": 495.49, "eval_steps_per_second": 7.949, "step": 6042 }, { "epoch": 58.0, "grad_norm": 0.6634971499443054, "learning_rate": 2.1e-05, "loss": 0.0396, "step": 6148 }, { "epoch": 58.0, "eval_LOCATION_f1": 0.8344370860927152, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8076923076923077, "eval_LOCATION_recall": 0.863013698630137, "eval_ORGANIZATION_f1": 0.7916666666666667, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.7215189873417721, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8996763754045306, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8742138364779874, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7142857142857144, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6097560975609756, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8055555555555555, "eval_TIME_number": 34, "eval_TIME_precision": 0.7631578947368421, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.16707535088062286, "eval_overall_accuracy": 0.9620969726802855, "eval_overall_f1": 0.839142091152815, "eval_overall_precision": 0.7924050632911392, "eval_overall_recall": 0.8917378917378918, "eval_runtime": 0.3757, "eval_samples_per_second": 497.712, "eval_steps_per_second": 7.985, "step": 6148 }, { "epoch": 59.0, "grad_norm": 0.4829442501068115, "learning_rate": 2.05e-05, "loss": 0.0392, "step": 6254 }, { "epoch": 59.0, "eval_LOCATION_f1": 0.8609271523178808, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8333333333333334, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7777777777777778, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.7088607594936709, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8888888888888887, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8717948717948718, "eval_PERSON_recall": 0.9066666666666666, "eval_QUANTITY_f1": 0.7761194029850746, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6842105263157895, "eval_QUANTITY_recall": 0.896551724137931, "eval_TIME_f1": 0.8285714285714286, "eval_TIME_number": 34, "eval_TIME_precision": 0.8055555555555556, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.15913118422031403, "eval_overall_accuracy": 0.9645582082205267, "eval_overall_f1": 0.8455284552845528, "eval_overall_precision": 0.8062015503875969, "eval_overall_recall": 0.8888888888888888, "eval_runtime": 0.3767, "eval_samples_per_second": 496.37, "eval_steps_per_second": 7.963, "step": 6254 }, { "epoch": 60.0, "grad_norm": 1.2957717180252075, "learning_rate": 2e-05, "loss": 0.0392, "step": 6360 }, { "epoch": 60.0, "eval_LOCATION_f1": 0.8609271523178808, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8333333333333334, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.763888888888889, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6962025316455697, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9055374592833875, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8853503184713376, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.746268656716418, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8285714285714286, "eval_TIME_number": 34, "eval_TIME_precision": 0.8055555555555556, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.16009843349456787, "eval_overall_accuracy": 0.9640659611124784, "eval_overall_f1": 0.8470906630581868, "eval_overall_precision": 0.8067010309278351, "eval_overall_recall": 0.8917378917378918, "eval_runtime": 0.3765, "eval_samples_per_second": 496.664, "eval_steps_per_second": 7.968, "step": 6360 }, { "epoch": 61.0, "grad_norm": 0.7019400000572205, "learning_rate": 1.9500000000000003e-05, "loss": 0.0396, "step": 6466 }, { "epoch": 61.0, "eval_LOCATION_f1": 0.8533333333333334, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8311688311688312, "eval_LOCATION_recall": 0.8767123287671232, "eval_ORGANIZATION_f1": 0.7581699346405228, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6590909090909091, "eval_ORGANIZATION_recall": 0.8923076923076924, "eval_PERSON_f1": 0.9090909090909091, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8860759493670886, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.625, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7999999999999999, "eval_TIME_number": 34, "eval_TIME_precision": 0.7777777777777778, "eval_TIME_recall": 0.8235294117647058, "eval_loss": 0.16811151802539825, "eval_overall_accuracy": 0.9620969726802855, "eval_overall_f1": 0.84, "eval_overall_precision": 0.7894736842105263, "eval_overall_recall": 0.8974358974358975, "eval_runtime": 0.3767, "eval_samples_per_second": 496.43, "eval_steps_per_second": 7.964, "step": 6466 }, { "epoch": 62.0, "grad_norm": 3.4165878295898438, "learning_rate": 1.9e-05, "loss": 0.0402, "step": 6572 }, { "epoch": 62.0, "eval_LOCATION_f1": 0.8344370860927152, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8076923076923077, "eval_LOCATION_recall": 0.863013698630137, "eval_ORGANIZATION_f1": 0.7534246575342466, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6790123456790124, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9120521172638437, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.89171974522293, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.7352941176470588, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6410256410256411, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7999999999999999, "eval_TIME_number": 34, "eval_TIME_precision": 0.7777777777777778, "eval_TIME_recall": 0.8235294117647058, "eval_loss": 0.16755960881710052, "eval_overall_accuracy": 0.9620969726802855, "eval_overall_f1": 0.8382749326145552, "eval_overall_precision": 0.7953964194373402, "eval_overall_recall": 0.886039886039886, "eval_runtime": 0.3754, "eval_samples_per_second": 498.126, "eval_steps_per_second": 7.991, "step": 6572 }, { "epoch": 63.0, "grad_norm": 1.2076839208602905, "learning_rate": 1.85e-05, "loss": 0.0391, "step": 6678 }, { "epoch": 63.0, "eval_LOCATION_f1": 0.8441558441558442, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8024691358024691, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7482993197278912, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6707317073170732, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8903225806451613, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8625, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7692307692307692, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6944444444444444, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8571428571428571, "eval_TIME_number": 34, "eval_TIME_precision": 0.8333333333333334, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.1606719195842743, "eval_overall_accuracy": 0.9640659611124784, "eval_overall_f1": 0.839142091152815, "eval_overall_precision": 0.7924050632911392, "eval_overall_recall": 0.8917378917378918, "eval_runtime": 0.3767, "eval_samples_per_second": 496.476, "eval_steps_per_second": 7.965, "step": 6678 }, { "epoch": 64.0, "grad_norm": 0.4508051872253418, "learning_rate": 1.8e-05, "loss": 0.0378, "step": 6784 }, { "epoch": 64.0, "eval_LOCATION_f1": 0.88, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8571428571428571, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.75, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6835443037974683, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.8903225806451613, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8625, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.746268656716418, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8405797101449276, "eval_TIME_number": 34, "eval_TIME_precision": 0.8285714285714286, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1592174917459488, "eval_overall_accuracy": 0.963081466896382, "eval_overall_f1": 0.8432432432432433, "eval_overall_precision": 0.8020565552699229, "eval_overall_recall": 0.8888888888888888, "eval_runtime": 0.3778, "eval_samples_per_second": 494.913, "eval_steps_per_second": 7.94, "step": 6784 }, { "epoch": 65.0, "grad_norm": 0.5349739193916321, "learning_rate": 1.75e-05, "loss": 0.0376, "step": 6890 }, { "epoch": 65.0, "eval_LOCATION_f1": 0.8496732026143791, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8125, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7586206896551724, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6875, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.9025974025974027, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.879746835443038, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.625, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7777777777777778, "eval_TIME_number": 34, "eval_TIME_precision": 0.7368421052631579, "eval_TIME_recall": 0.8235294117647058, "eval_loss": 0.16776202619075775, "eval_overall_accuracy": 0.9618508491262614, "eval_overall_f1": 0.8353413654618473, "eval_overall_precision": 0.7878787878787878, "eval_overall_recall": 0.8888888888888888, "eval_runtime": 0.3743, "eval_samples_per_second": 499.657, "eval_steps_per_second": 8.016, "step": 6890 }, { "epoch": 66.0, "grad_norm": 0.8931418657302856, "learning_rate": 1.7000000000000003e-05, "loss": 0.0369, "step": 6996 }, { "epoch": 66.0, "eval_LOCATION_f1": 0.8684210526315789, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8354430379746836, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.7450980392156863, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6477272727272727, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8932038834951457, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8679245283018868, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7352941176470588, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6410256410256411, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.16921161115169525, "eval_overall_accuracy": 0.9606202313561408, "eval_overall_f1": 0.8366533864541833, "eval_overall_precision": 0.7835820895522388, "eval_overall_recall": 0.8974358974358975, "eval_runtime": 0.3803, "eval_samples_per_second": 491.763, "eval_steps_per_second": 7.889, "step": 6996 }, { "epoch": 67.0, "grad_norm": 0.1230609193444252, "learning_rate": 1.65e-05, "loss": 0.0366, "step": 7102 }, { "epoch": 67.0, "eval_LOCATION_f1": 0.8496732026143791, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8125, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7724137931034483, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.7, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.896103896103896, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8734177215189873, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.625, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8285714285714286, "eval_TIME_number": 34, "eval_TIME_precision": 0.8055555555555556, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1675971895456314, "eval_overall_accuracy": 0.9628353433423579, "eval_overall_f1": 0.8402684563758389, "eval_overall_precision": 0.7944162436548223, "eval_overall_recall": 0.8917378917378918, "eval_runtime": 0.3749, "eval_samples_per_second": 498.851, "eval_steps_per_second": 8.003, "step": 7102 }, { "epoch": 68.0, "grad_norm": 0.18901905417442322, "learning_rate": 1.6000000000000003e-05, "loss": 0.0353, "step": 7208 }, { "epoch": 68.0, "eval_LOCATION_f1": 0.8609271523178808, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8333333333333334, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7215189873417722, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6129032258064516, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8903225806451613, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8625, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.625, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1709582656621933, "eval_overall_accuracy": 0.9596357371400444, "eval_overall_f1": 0.8274044795783927, "eval_overall_precision": 0.7696078431372549, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3746, "eval_samples_per_second": 499.264, "eval_steps_per_second": 8.01, "step": 7208 }, { "epoch": 69.0, "grad_norm": 0.19151590764522552, "learning_rate": 1.55e-05, "loss": 0.0342, "step": 7314 }, { "epoch": 69.0, "eval_LOCATION_f1": 0.8609271523178808, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8333333333333334, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7417218543046358, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6511627906976745, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8996763754045306, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8742138364779874, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7575757575757576, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8285714285714286, "eval_TIME_number": 34, "eval_TIME_precision": 0.8055555555555556, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.16525578498840332, "eval_overall_accuracy": 0.9645582082205267, "eval_overall_f1": 0.8406961178045514, "eval_overall_precision": 0.7929292929292929, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3767, "eval_samples_per_second": 496.387, "eval_steps_per_second": 7.963, "step": 7314 }, { "epoch": 70.0, "grad_norm": 0.4046976864337921, "learning_rate": 1.5e-05, "loss": 0.0339, "step": 7420 }, { "epoch": 70.0, "eval_LOCATION_f1": 0.8627450980392157, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.825, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.7586206896551724, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6875, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8996763754045306, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8742138364779874, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7761194029850746, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6842105263157895, "eval_QUANTITY_recall": 0.896551724137931, "eval_TIME_f1": 0.8285714285714286, "eval_TIME_number": 34, "eval_TIME_precision": 0.8055555555555556, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.16819427907466888, "eval_overall_accuracy": 0.9640659611124784, "eval_overall_f1": 0.8467741935483871, "eval_overall_precision": 0.8015267175572519, "eval_overall_recall": 0.8974358974358975, "eval_runtime": 0.3742, "eval_samples_per_second": 499.708, "eval_steps_per_second": 8.017, "step": 7420 }, { "epoch": 71.0, "grad_norm": 0.9475017189979553, "learning_rate": 1.45e-05, "loss": 0.0358, "step": 7526 }, { "epoch": 71.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.75, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6551724137931034, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8938906752411575, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8633540372670807, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.6956521739130435, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.7999999999999999, "eval_TIME_number": 34, "eval_TIME_precision": 0.7777777777777778, "eval_TIME_recall": 0.8235294117647058, "eval_loss": 0.1757313460111618, "eval_overall_accuracy": 0.9616047255722373, "eval_overall_f1": 0.8324468085106382, "eval_overall_precision": 0.7805486284289277, "eval_overall_recall": 0.8917378917378918, "eval_runtime": 0.3769, "eval_samples_per_second": 496.1, "eval_steps_per_second": 7.959, "step": 7526 }, { "epoch": 72.0, "grad_norm": 0.22020655870437622, "learning_rate": 1.4000000000000001e-05, "loss": 0.0333, "step": 7632 }, { "epoch": 72.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7534246575342466, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6790123456790124, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.896103896103896, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8734177215189873, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7272727272727273, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6486486486486487, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8285714285714286, "eval_TIME_number": 34, "eval_TIME_precision": 0.8055555555555556, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.17391657829284668, "eval_overall_accuracy": 0.9638198375584544, "eval_overall_f1": 0.8405405405405405, "eval_overall_precision": 0.7994858611825193, "eval_overall_recall": 0.886039886039886, "eval_runtime": 0.3764, "eval_samples_per_second": 496.821, "eval_steps_per_second": 7.97, "step": 7632 }, { "epoch": 73.0, "grad_norm": 0.4557439088821411, "learning_rate": 1.3500000000000001e-05, "loss": 0.0346, "step": 7738 }, { "epoch": 73.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7777777777777778, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.7088607594936709, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9061488673139159, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8805031446540881, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.75, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6857142857142857, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8115942028985507, "eval_TIME_number": 34, "eval_TIME_precision": 0.8, "eval_TIME_recall": 0.8235294117647058, "eval_loss": 0.1734769642353058, "eval_overall_accuracy": 0.9655427024366232, "eval_overall_f1": 0.8505434782608695, "eval_overall_precision": 0.812987012987013, "eval_overall_recall": 0.8917378917378918, "eval_runtime": 0.3768, "eval_samples_per_second": 496.296, "eval_steps_per_second": 7.962, "step": 7738 }, { "epoch": 74.0, "grad_norm": 0.21247869729995728, "learning_rate": 1.3000000000000001e-05, "loss": 0.033, "step": 7844 }, { "epoch": 74.0, "eval_LOCATION_f1": 0.8609271523178808, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8333333333333334, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7517730496453899, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6973684210526315, "eval_ORGANIZATION_recall": 0.8153846153846154, "eval_PERSON_f1": 0.8938906752411575, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8633540372670807, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7058823529411765, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6153846153846154, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8285714285714286, "eval_TIME_number": 34, "eval_TIME_precision": 0.8055555555555556, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1750202625989914, "eval_overall_accuracy": 0.9633275904504061, "eval_overall_f1": 0.8367071524966261, "eval_overall_precision": 0.7948717948717948, "eval_overall_recall": 0.8831908831908832, "eval_runtime": 0.375, "eval_samples_per_second": 498.628, "eval_steps_per_second": 7.999, "step": 7844 }, { "epoch": 75.0, "grad_norm": 0.42767712473869324, "learning_rate": 1.25e-05, "loss": 0.0331, "step": 7950 }, { "epoch": 75.0, "eval_LOCATION_f1": 0.8741721854304636, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8461538461538461, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.7702702702702702, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6867469879518072, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.9019607843137256, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8846153846153846, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7352941176470588, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6410256410256411, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8285714285714286, "eval_TIME_number": 34, "eval_TIME_precision": 0.8055555555555556, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.17306743562221527, "eval_overall_accuracy": 0.9628353433423579, "eval_overall_f1": 0.847913862718708, "eval_overall_precision": 0.8035714285714286, "eval_overall_recall": 0.8974358974358975, "eval_runtime": 0.377, "eval_samples_per_second": 496.081, "eval_steps_per_second": 7.959, "step": 7950 }, { "epoch": 76.0, "grad_norm": 0.09305271506309509, "learning_rate": 1.2e-05, "loss": 0.0342, "step": 8056 }, { "epoch": 76.0, "eval_LOCATION_f1": 0.8627450980392157, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.825, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.7567567567567568, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6746987951807228, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9025974025974027, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.879746835443038, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7384615384615385, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6666666666666666, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8405797101449276, "eval_TIME_number": 34, "eval_TIME_precision": 0.8285714285714286, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.17322573065757751, "eval_overall_accuracy": 0.963081466896382, "eval_overall_f1": 0.8452220726783309, "eval_overall_precision": 0.8010204081632653, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3744, "eval_samples_per_second": 499.512, "eval_steps_per_second": 8.014, "step": 8056 }, { "epoch": 77.0, "grad_norm": 0.8506590127944946, "learning_rate": 1.1500000000000002e-05, "loss": 0.0327, "step": 8162 }, { "epoch": 77.0, "eval_LOCATION_f1": 0.8741721854304636, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8461538461538461, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.7808219178082192, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.7037037037037037, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8932038834951457, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8679245283018868, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7812500000000001, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.7142857142857143, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8695652173913043, "eval_TIME_number": 34, "eval_TIME_precision": 0.8571428571428571, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.17164190113544464, "eval_overall_accuracy": 0.9660349495446714, "eval_overall_f1": 0.8552097428958052, "eval_overall_precision": 0.8144329896907216, "eval_overall_recall": 0.9002849002849003, "eval_runtime": 0.375, "eval_samples_per_second": 498.614, "eval_steps_per_second": 7.999, "step": 8162 }, { "epoch": 78.0, "grad_norm": 0.9322987794876099, "learning_rate": 1.1000000000000001e-05, "loss": 0.0327, "step": 8268 }, { "epoch": 78.0, "eval_LOCATION_f1": 0.8627450980392157, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.825, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.7297297297297298, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6506024096385542, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.8990228013029316, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8789808917197452, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7352941176470588, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6410256410256411, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8529411764705882, "eval_TIME_number": 34, "eval_TIME_precision": 0.8529411764705882, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1736339032649994, "eval_overall_accuracy": 0.963081466896382, "eval_overall_f1": 0.8387096774193549, "eval_overall_precision": 0.7938931297709924, "eval_overall_recall": 0.8888888888888888, "eval_runtime": 0.3754, "eval_samples_per_second": 498.099, "eval_steps_per_second": 7.991, "step": 8268 }, { "epoch": 79.0, "grad_norm": 0.6149049401283264, "learning_rate": 1.05e-05, "loss": 0.0302, "step": 8374 }, { "epoch": 79.0, "eval_LOCATION_f1": 0.8684210526315789, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8354430379746836, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.7567567567567568, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6746987951807228, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.8932038834951457, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8679245283018868, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7692307692307692, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6944444444444444, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.7887323943661971, "eval_TIME_number": 34, "eval_TIME_precision": 0.7567567567567568, "eval_TIME_recall": 0.8235294117647058, "eval_loss": 0.17478498816490173, "eval_overall_accuracy": 0.9633275904504061, "eval_overall_f1": 0.8402684563758389, "eval_overall_precision": 0.7944162436548223, "eval_overall_recall": 0.8917378917378918, "eval_runtime": 0.3751, "eval_samples_per_second": 498.475, "eval_steps_per_second": 7.997, "step": 8374 }, { "epoch": 80.0, "grad_norm": 0.601209282875061, "learning_rate": 1e-05, "loss": 0.0299, "step": 8480 }, { "epoch": 80.0, "eval_LOCATION_f1": 0.8741721854304636, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8461538461538461, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.7972972972972974, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.7108433734939759, "eval_ORGANIZATION_recall": 0.9076923076923077, "eval_PERSON_f1": 0.8990228013029316, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8789808917197452, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7692307692307692, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6944444444444444, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8571428571428571, "eval_TIME_number": 34, "eval_TIME_precision": 0.8333333333333334, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.1755015254020691, "eval_overall_accuracy": 0.9645582082205267, "eval_overall_f1": 0.8582995951417003, "eval_overall_precision": 0.8153846153846154, "eval_overall_recall": 0.905982905982906, "eval_runtime": 0.377, "eval_samples_per_second": 496.085, "eval_steps_per_second": 7.959, "step": 8480 }, { "epoch": 81.0, "grad_norm": 0.7238019704818726, "learning_rate": 9.5e-06, "loss": 0.0316, "step": 8586 }, { "epoch": 81.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7482993197278912, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6707317073170732, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8859934853420196, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8662420382165605, "eval_PERSON_recall": 0.9066666666666666, "eval_QUANTITY_f1": 0.7272727272727273, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6486486486486487, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8405797101449276, "eval_TIME_number": 34, "eval_TIME_precision": 0.8285714285714286, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.17351825535297394, "eval_overall_accuracy": 0.9633275904504061, "eval_overall_f1": 0.8362652232746955, "eval_overall_precision": 0.7963917525773195, "eval_overall_recall": 0.8803418803418803, "eval_runtime": 0.3787, "eval_samples_per_second": 493.799, "eval_steps_per_second": 7.922, "step": 8586 }, { "epoch": 82.0, "grad_norm": 0.5620818138122559, "learning_rate": 9e-06, "loss": 0.0294, "step": 8692 }, { "epoch": 82.0, "eval_LOCATION_f1": 0.8684210526315789, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8354430379746836, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.7482993197278912, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6707317073170732, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.896774193548387, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.86875, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7575757575757576, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8656716417910447, "eval_TIME_number": 34, "eval_TIME_precision": 0.8787878787878788, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.17314855754375458, "eval_overall_accuracy": 0.9633275904504061, "eval_overall_f1": 0.8463611859838276, "eval_overall_precision": 0.80306905370844, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3766, "eval_samples_per_second": 496.551, "eval_steps_per_second": 7.966, "step": 8692 }, { "epoch": 83.0, "grad_norm": 0.18742328882217407, "learning_rate": 8.500000000000002e-06, "loss": 0.0303, "step": 8798 }, { "epoch": 83.0, "eval_LOCATION_f1": 0.8533333333333334, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8311688311688312, "eval_LOCATION_recall": 0.8767123287671232, "eval_ORGANIZATION_f1": 0.7482993197278912, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6707317073170732, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8932038834951457, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8679245283018868, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.6956521739130435, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8055555555555555, "eval_TIME_number": 34, "eval_TIME_precision": 0.7631578947368421, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1789408028125763, "eval_overall_accuracy": 0.9618508491262614, "eval_overall_f1": 0.8299866131191433, "eval_overall_precision": 0.7828282828282829, "eval_overall_recall": 0.8831908831908832, "eval_runtime": 0.384, "eval_samples_per_second": 487.03, "eval_steps_per_second": 7.813, "step": 8798 }, { "epoch": 84.0, "grad_norm": 0.7289339900016785, "learning_rate": 8.000000000000001e-06, "loss": 0.0299, "step": 8904 }, { "epoch": 84.0, "eval_LOCATION_f1": 0.8609271523178808, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8333333333333334, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7534246575342466, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6790123456790124, "eval_ORGANIZATION_recall": 0.8461538461538461, "eval_PERSON_f1": 0.8938906752411575, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8633540372670807, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7575757575757576, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8450704225352113, "eval_TIME_number": 34, "eval_TIME_precision": 0.8108108108108109, "eval_TIME_recall": 0.8823529411764706, "eval_loss": 0.17691366374492645, "eval_overall_accuracy": 0.9638198375584544, "eval_overall_f1": 0.8429530201342281, "eval_overall_precision": 0.7969543147208121, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3782, "eval_samples_per_second": 494.403, "eval_steps_per_second": 7.932, "step": 8904 }, { "epoch": 85.0, "grad_norm": 2.1489155292510986, "learning_rate": 7.5e-06, "loss": 0.031, "step": 9010 }, { "epoch": 85.0, "eval_LOCATION_f1": 0.8609271523178808, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8333333333333334, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7763157894736842, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6781609195402298, "eval_ORGANIZATION_recall": 0.9076923076923077, "eval_PERSON_f1": 0.9055374592833875, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8853503184713376, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7575757575757576, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8055555555555555, "eval_TIME_number": 34, "eval_TIME_precision": 0.7631578947368421, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.18658407032489777, "eval_overall_accuracy": 0.9620969726802855, "eval_overall_f1": 0.8475935828877006, "eval_overall_precision": 0.7984886649874056, "eval_overall_recall": 0.9031339031339032, "eval_runtime": 0.3791, "eval_samples_per_second": 493.321, "eval_steps_per_second": 7.914, "step": 9010 }, { "epoch": 86.0, "grad_norm": 0.6923357248306274, "learning_rate": 7.000000000000001e-06, "loss": 0.0303, "step": 9116 }, { "epoch": 86.0, "eval_LOCATION_f1": 0.8456375838926175, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8289473684210527, "eval_LOCATION_recall": 0.863013698630137, "eval_ORGANIZATION_f1": 0.7552447552447553, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6923076923076923, "eval_ORGANIZATION_recall": 0.8307692307692308, "eval_PERSON_f1": 0.8938906752411575, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8633540372670807, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.746268656716418, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8285714285714286, "eval_TIME_number": 34, "eval_TIME_precision": 0.8055555555555556, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1748000532388687, "eval_overall_accuracy": 0.9628353433423579, "eval_overall_f1": 0.8378378378378378, "eval_overall_precision": 0.7969151670951157, "eval_overall_recall": 0.8831908831908832, "eval_runtime": 0.3795, "eval_samples_per_second": 492.769, "eval_steps_per_second": 7.905, "step": 9116 }, { "epoch": 87.0, "grad_norm": 2.244084119796753, "learning_rate": 6.5000000000000004e-06, "loss": 0.0302, "step": 9222 }, { "epoch": 87.0, "eval_LOCATION_f1": 0.8741721854304636, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8461538461538461, "eval_LOCATION_recall": 0.9041095890410958, "eval_ORGANIZATION_f1": 0.7702702702702702, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6867469879518072, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8990228013029316, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8789808917197452, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7575757575757576, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8358208955223881, "eval_TIME_number": 34, "eval_TIME_precision": 0.8484848484848485, "eval_TIME_recall": 0.8235294117647058, "eval_loss": 0.17357194423675537, "eval_overall_accuracy": 0.9643120846665025, "eval_overall_f1": 0.8497970230040596, "eval_overall_precision": 0.8092783505154639, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3746, "eval_samples_per_second": 499.189, "eval_steps_per_second": 8.008, "step": 9222 }, { "epoch": 88.0, "grad_norm": 0.33963289856910706, "learning_rate": 6e-06, "loss": 0.0294, "step": 9328 }, { "epoch": 88.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7755102040816327, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6951219512195121, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.9055374592833875, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8853503184713376, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.75, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6857142857142857, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8405797101449276, "eval_TIME_number": 34, "eval_TIME_precision": 0.8285714285714286, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.17453482747077942, "eval_overall_accuracy": 0.965296578882599, "eval_overall_f1": 0.8521031207598372, "eval_overall_precision": 0.8134715025906736, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3757, "eval_samples_per_second": 497.729, "eval_steps_per_second": 7.985, "step": 9328 }, { "epoch": 89.0, "grad_norm": 0.7375528216362, "learning_rate": 5.500000000000001e-06, "loss": 0.028, "step": 9434 }, { "epoch": 89.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7671232876712328, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.691358024691358, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9090909090909091, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8860759493670886, "eval_PERSON_recall": 0.9333333333333333, "eval_QUANTITY_f1": 0.716417910447761, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.631578947368421, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8285714285714286, "eval_TIME_number": 34, "eval_TIME_precision": 0.8055555555555556, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.17815575003623962, "eval_overall_accuracy": 0.9635737140044303, "eval_overall_f1": 0.8475033738191633, "eval_overall_precision": 0.8051282051282052, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3755, "eval_samples_per_second": 498.03, "eval_steps_per_second": 7.99, "step": 9434 }, { "epoch": 90.0, "grad_norm": 0.9239028096199036, "learning_rate": 5e-06, "loss": 0.031, "step": 9540 }, { "epoch": 90.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7619047619047619, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6829268292682927, "eval_ORGANIZATION_recall": 0.8615384615384616, "eval_PERSON_f1": 0.9025974025974027, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.879746835443038, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.716417910447761, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.631578947368421, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8405797101449276, "eval_TIME_number": 34, "eval_TIME_precision": 0.8285714285714286, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.17805850505828857, "eval_overall_accuracy": 0.9635737140044303, "eval_overall_f1": 0.844804318488529, "eval_overall_precision": 0.8025641025641026, "eval_overall_recall": 0.8917378917378918, "eval_runtime": 0.3757, "eval_samples_per_second": 497.747, "eval_steps_per_second": 7.985, "step": 9540 }, { "epoch": 91.0, "grad_norm": 0.8207985758781433, "learning_rate": 4.5e-06, "loss": 0.0296, "step": 9646 }, { "epoch": 91.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.76, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6705882352941176, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.9025974025974027, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.879746835443038, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7058823529411765, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6153846153846154, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8285714285714286, "eval_TIME_number": 34, "eval_TIME_precision": 0.8055555555555556, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.18207119405269623, "eval_overall_accuracy": 0.9633275904504061, "eval_overall_f1": 0.8418230563002682, "eval_overall_precision": 0.7949367088607595, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3764, "eval_samples_per_second": 496.87, "eval_steps_per_second": 7.971, "step": 9646 }, { "epoch": 92.0, "grad_norm": 0.7969566583633423, "learning_rate": 4.000000000000001e-06, "loss": 0.0293, "step": 9752 }, { "epoch": 92.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7651006711409396, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6785714285714286, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.9055374592833875, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8853503184713376, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.716417910447761, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.631578947368421, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.18025565147399902, "eval_overall_accuracy": 0.9633275904504061, "eval_overall_f1": 0.8440860215053764, "eval_overall_precision": 0.7989821882951654, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3775, "eval_samples_per_second": 495.387, "eval_steps_per_second": 7.947, "step": 9752 }, { "epoch": 93.0, "grad_norm": 0.06257300078868866, "learning_rate": 3.5000000000000004e-06, "loss": 0.0263, "step": 9858 }, { "epoch": 93.0, "eval_LOCATION_f1": 0.8609271523178808, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8333333333333334, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7651006711409396, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6785714285714286, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.896103896103896, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8734177215189873, "eval_PERSON_recall": 0.92, "eval_QUANTITY_f1": 0.7575757575757576, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.8620689655172413, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.18161390721797943, "eval_overall_accuracy": 0.9628353433423579, "eval_overall_f1": 0.8429530201342281, "eval_overall_precision": 0.7969543147208121, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3753, "eval_samples_per_second": 498.313, "eval_steps_per_second": 7.994, "step": 9858 }, { "epoch": 94.0, "grad_norm": 0.7405329942703247, "learning_rate": 3e-06, "loss": 0.0266, "step": 9964 }, { "epoch": 94.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7919463087248322, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.7023809523809523, "eval_ORGANIZATION_recall": 0.9076923076923077, "eval_PERSON_f1": 0.8996763754045306, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8742138364779874, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7272727272727273, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6486486486486487, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.18204358220100403, "eval_overall_accuracy": 0.963081466896382, "eval_overall_f1": 0.8483221476510067, "eval_overall_precision": 0.8020304568527918, "eval_overall_recall": 0.9002849002849003, "eval_runtime": 0.3753, "eval_samples_per_second": 498.26, "eval_steps_per_second": 7.993, "step": 9964 }, { "epoch": 95.0, "grad_norm": 0.20907868444919586, "learning_rate": 2.5e-06, "loss": 0.0277, "step": 10070 }, { "epoch": 95.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7866666666666665, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6941176470588235, "eval_ORGANIZATION_recall": 0.9076923076923077, "eval_PERSON_f1": 0.8996763754045306, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8742138364779874, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.716417910447761, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.631578947368421, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.18414747714996338, "eval_overall_accuracy": 0.963081466896382, "eval_overall_f1": 0.8460508701472558, "eval_overall_precision": 0.797979797979798, "eval_overall_recall": 0.9002849002849003, "eval_runtime": 0.3755, "eval_samples_per_second": 498.017, "eval_steps_per_second": 7.99, "step": 10070 }, { "epoch": 96.0, "grad_norm": 0.9583673477172852, "learning_rate": 2.0000000000000003e-06, "loss": 0.0273, "step": 10176 }, { "epoch": 96.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7702702702702702, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6867469879518072, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8996763754045306, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8742138364779874, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.716417910447761, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.631578947368421, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.1805674433708191, "eval_overall_accuracy": 0.9633275904504061, "eval_overall_f1": 0.8429530201342281, "eval_overall_precision": 0.7969543147208121, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3752, "eval_samples_per_second": 498.391, "eval_steps_per_second": 7.996, "step": 10176 }, { "epoch": 97.0, "grad_norm": 0.43106725811958313, "learning_rate": 1.5e-06, "loss": 0.027, "step": 10282 }, { "epoch": 97.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.76, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6705882352941176, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8996763754045306, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8742138364779874, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.7058823529411765, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.6153846153846154, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.18312983214855194, "eval_overall_accuracy": 0.9620969726802855, "eval_overall_f1": 0.839572192513369, "eval_overall_precision": 0.7909319899244333, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3786, "eval_samples_per_second": 493.934, "eval_steps_per_second": 7.924, "step": 10282 }, { "epoch": 98.0, "grad_norm": 0.7926185131072998, "learning_rate": 1.0000000000000002e-06, "loss": 0.0266, "step": 10388 }, { "epoch": 98.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7837837837837838, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6987951807228916, "eval_ORGANIZATION_recall": 0.8923076923076924, "eval_PERSON_f1": 0.8996763754045306, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8742138364779874, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.716417910447761, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.631578947368421, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.18297931551933289, "eval_overall_accuracy": 0.9628353433423579, "eval_overall_f1": 0.8456375838926175, "eval_overall_precision": 0.799492385786802, "eval_overall_recall": 0.8974358974358975, "eval_runtime": 0.3765, "eval_samples_per_second": 496.713, "eval_steps_per_second": 7.969, "step": 10388 }, { "epoch": 99.0, "grad_norm": 0.46750637888908386, "learning_rate": 5.000000000000001e-07, "loss": 0.0273, "step": 10494 }, { "epoch": 99.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7837837837837838, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6987951807228916, "eval_ORGANIZATION_recall": 0.8923076923076924, "eval_PERSON_f1": 0.8996763754045306, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8742138364779874, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.716417910447761, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.631578947368421, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.18302349746227264, "eval_overall_accuracy": 0.9628353433423579, "eval_overall_f1": 0.8456375838926175, "eval_overall_precision": 0.799492385786802, "eval_overall_recall": 0.8974358974358975, "eval_runtime": 0.3764, "eval_samples_per_second": 496.813, "eval_steps_per_second": 7.97, "step": 10494 }, { "epoch": 100.0, "grad_norm": 0.4663177728652954, "learning_rate": 0.0, "loss": 0.0281, "step": 10600 }, { "epoch": 100.0, "eval_LOCATION_f1": 0.8666666666666666, "eval_LOCATION_number": 73, "eval_LOCATION_precision": 0.8441558441558441, "eval_LOCATION_recall": 0.8904109589041096, "eval_ORGANIZATION_f1": 0.7651006711409396, "eval_ORGANIZATION_number": 65, "eval_ORGANIZATION_precision": 0.6785714285714286, "eval_ORGANIZATION_recall": 0.8769230769230769, "eval_PERSON_f1": 0.8996763754045306, "eval_PERSON_number": 150, "eval_PERSON_precision": 0.8742138364779874, "eval_PERSON_recall": 0.9266666666666666, "eval_QUANTITY_f1": 0.716417910447761, "eval_QUANTITY_number": 29, "eval_QUANTITY_precision": 0.631578947368421, "eval_QUANTITY_recall": 0.8275862068965517, "eval_TIME_f1": 0.8169014084507041, "eval_TIME_number": 34, "eval_TIME_precision": 0.7837837837837838, "eval_TIME_recall": 0.8529411764705882, "eval_loss": 0.18232890963554382, "eval_overall_accuracy": 0.9625892197883338, "eval_overall_f1": 0.8418230563002682, "eval_overall_precision": 0.7949367088607595, "eval_overall_recall": 0.8945868945868946, "eval_runtime": 0.3776, "eval_samples_per_second": 495.236, "eval_steps_per_second": 7.945, "step": 10600 }, { "epoch": 100.0, "step": 10600, "total_flos": 5043631411410816.0, "train_loss": 0.07345248532745073, "train_runtime": 628.8015, "train_samples_per_second": 268.447, "train_steps_per_second": 16.857 } ], "logging_steps": 500, "max_steps": 10600, "num_input_tokens_seen": 0, "num_train_epochs": 100, "save_steps": 500, "total_flos": 5043631411410816.0, "train_batch_size": 16, "trial_name": null, "trial_params": null }