diff --git "a/trainer_state.json" "b/trainer_state.json" --- "a/trainer_state.json" +++ "b/trainer_state.json" @@ -10,9 +10,9 @@ "log_history": [ { "epoch": 1.0, - "grad_norm": 1.1694915294647217, + "grad_norm": 1.2873680591583252, "learning_rate": 4.9500000000000004e-05, - "loss": 1.2426, + "loss": 1.247, "step": 106 }, { @@ -37,21 +37,21 @@ "eval_TIME_number": 31, "eval_TIME_precision": 0.0, "eval_TIME_recall": 0.0, - "eval_loss": 0.7388785481452942, + "eval_loss": 0.7410362362861633, "eval_overall_accuracy": 0.8352737181075056, "eval_overall_f1": 0.0, "eval_overall_precision": 0.0, "eval_overall_recall": 0.0, - "eval_runtime": 0.9307, - "eval_samples_per_second": 200.92, - "eval_steps_per_second": 3.223, + "eval_runtime": 0.371, + "eval_samples_per_second": 504.108, + "eval_steps_per_second": 8.087, "step": 106 }, { "epoch": 2.0, - "grad_norm": 1.3079735040664673, + "grad_norm": 1.251749873161316, "learning_rate": 4.9e-05, - "loss": 0.7083, + "loss": 0.7131, "step": 212 }, { @@ -64,9 +64,9 @@ "eval_ORGANIZATION_number": 61, "eval_ORGANIZATION_precision": 0.0, "eval_ORGANIZATION_recall": 0.0, - "eval_PERSON_f1": 0.012658227848101266, + "eval_PERSON_f1": 0.012738853503184716, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.2, + "eval_PERSON_precision": 0.25, "eval_PERSON_recall": 0.006535947712418301, "eval_QUANTITY_f1": 0.0, "eval_QUANTITY_number": 32, @@ -76,21 +76,21 @@ "eval_TIME_number": 31, "eval_TIME_precision": 0.0, "eval_TIME_recall": 0.0, - "eval_loss": 0.6314201951026917, + "eval_loss": 0.6369233727455139, "eval_overall_accuracy": 0.8362645528858063, - "eval_overall_f1": 0.0056657223796034, - "eval_overall_precision": 0.14285714285714285, + "eval_overall_f1": 0.005681818181818181, + "eval_overall_precision": 0.16666666666666666, "eval_overall_recall": 0.002890173410404624, - "eval_runtime": 0.9862, - "eval_samples_per_second": 189.623, - "eval_steps_per_second": 3.042, + "eval_runtime": 0.3741, + "eval_samples_per_second": 499.916, + "eval_steps_per_second": 8.02, "step": 212 }, { "epoch": 3.0, - "grad_norm": 1.4049551486968994, + "grad_norm": 1.4780464172363281, "learning_rate": 4.85e-05, - "loss": 0.6256, + "loss": 0.629, "step": 318 }, { @@ -103,10 +103,10 @@ "eval_ORGANIZATION_number": 61, "eval_ORGANIZATION_precision": 0.0, "eval_ORGANIZATION_recall": 0.0, - "eval_PERSON_f1": 0.06779661016949153, + "eval_PERSON_f1": 0.07954545454545454, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.25, - "eval_PERSON_recall": 0.0392156862745098, + "eval_PERSON_precision": 0.30434782608695654, + "eval_PERSON_recall": 0.0457516339869281, "eval_QUANTITY_f1": 0.0, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.0, @@ -115,37 +115,37 @@ "eval_TIME_number": 31, "eval_TIME_precision": 0.0, "eval_TIME_recall": 0.0, - "eval_loss": 0.534466028213501, - "eval_overall_accuracy": 0.8429526876393362, - "eval_overall_f1": 0.031331592689295036, - "eval_overall_precision": 0.16216216216216217, - "eval_overall_recall": 0.017341040462427744, - "eval_runtime": 1.0086, - "eval_samples_per_second": 185.399, - "eval_steps_per_second": 2.974, + "eval_loss": 0.5407068729400635, + "eval_overall_accuracy": 0.8424572702501858, + "eval_overall_f1": 0.03674540682414699, + "eval_overall_precision": 0.2, + "eval_overall_recall": 0.02023121387283237, + "eval_runtime": 0.3721, + "eval_samples_per_second": 502.555, + "eval_steps_per_second": 8.062, "step": 318 }, { "epoch": 4.0, - "grad_norm": 2.0670742988586426, + "grad_norm": 2.4032466411590576, "learning_rate": 4.8e-05, - "loss": 0.5318, + "loss": 0.5381, "step": 424 }, { "epoch": 4.0, - "eval_LOCATION_f1": 0.15384615384615383, + "eval_LOCATION_f1": 0.10389610389610389, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6666666666666666, - "eval_LOCATION_recall": 0.08695652173913043, - "eval_ORGANIZATION_f1": 0.02777777777777778, + "eval_LOCATION_precision": 0.5, + "eval_LOCATION_recall": 0.057971014492753624, + "eval_ORGANIZATION_f1": 0.028169014084507043, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.09090909090909091, + "eval_ORGANIZATION_precision": 0.1, "eval_ORGANIZATION_recall": 0.01639344262295082, - "eval_PERSON_f1": 0.27999999999999997, + "eval_PERSON_f1": 0.23236514522821577, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.36082474226804123, - "eval_PERSON_recall": 0.22875816993464052, + "eval_PERSON_precision": 0.3181818181818182, + "eval_PERSON_recall": 0.1830065359477124, "eval_QUANTITY_f1": 0.0, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.0, @@ -154,1113 +154,1113 @@ "eval_TIME_number": 31, "eval_TIME_precision": 0.3888888888888889, "eval_TIME_recall": 0.22580645161290322, - "eval_loss": 0.4316946566104889, - "eval_overall_accuracy": 0.8704483527371811, - "eval_overall_f1": 0.20374220374220375, - "eval_overall_precision": 0.362962962962963, - "eval_overall_recall": 0.1416184971098266, - "eval_runtime": 0.9805, - "eval_samples_per_second": 190.727, - "eval_steps_per_second": 3.06, + "eval_loss": 0.4457257091999054, + "eval_overall_accuracy": 0.8667327223185534, + "eval_overall_f1": 0.1702127659574468, + "eval_overall_precision": 0.3225806451612903, + "eval_overall_recall": 0.11560693641618497, + "eval_runtime": 0.3729, + "eval_samples_per_second": 501.455, + "eval_steps_per_second": 8.045, "step": 424 }, { "epoch": 5.0, - "grad_norm": 1.8049697875976562, + "grad_norm": 1.463838815689087, "learning_rate": 4.75e-05, - "loss": 0.4253, + "loss": 0.4387, "step": 530 }, { "epoch": 5.0, - "eval_LOCATION_f1": 0.3669724770642202, + "eval_LOCATION_f1": 0.34951456310679613, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.5, - "eval_LOCATION_recall": 0.2898550724637681, - "eval_ORGANIZATION_f1": 0.22222222222222224, + "eval_LOCATION_precision": 0.5294117647058824, + "eval_LOCATION_recall": 0.2608695652173913, + "eval_ORGANIZATION_f1": 0.16470588235294117, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.3448275862068966, - "eval_ORGANIZATION_recall": 0.16393442622950818, - "eval_PERSON_f1": 0.6455696202531646, + "eval_ORGANIZATION_precision": 0.2916666666666667, + "eval_ORGANIZATION_recall": 0.11475409836065574, + "eval_PERSON_f1": 0.5243902439024392, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.6257668711656442, - "eval_PERSON_recall": 0.6666666666666666, + "eval_PERSON_precision": 0.49142857142857144, + "eval_PERSON_recall": 0.5620915032679739, "eval_QUANTITY_f1": 0.0, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.0, "eval_QUANTITY_recall": 0.0, - "eval_TIME_f1": 0.6129032258064516, - "eval_TIME_number": 31, - "eval_TIME_precision": 0.6129032258064516, - "eval_TIME_recall": 0.6129032258064516, - "eval_loss": 0.3396449685096741, - "eval_overall_accuracy": 0.9061184047560069, - "eval_overall_f1": 0.4878836833602585, - "eval_overall_precision": 0.5531135531135531, - "eval_overall_recall": 0.43641618497109824, - "eval_runtime": 0.9459, - "eval_samples_per_second": 197.69, - "eval_steps_per_second": 3.171, + "eval_TIME_f1": 0.45283018867924524, + "eval_TIME_number": 31, + "eval_TIME_precision": 0.5454545454545454, + "eval_TIME_recall": 0.3870967741935484, + "eval_loss": 0.3568115234375, + "eval_overall_accuracy": 0.8957146395838494, + "eval_overall_f1": 0.4026186579378069, + "eval_overall_precision": 0.4641509433962264, + "eval_overall_recall": 0.3554913294797688, + "eval_runtime": 0.3737, + "eval_samples_per_second": 500.37, + "eval_steps_per_second": 8.027, "step": 530 }, { "epoch": 6.0, - "grad_norm": 1.4327764511108398, + "grad_norm": 9.46357250213623, "learning_rate": 4.7e-05, - "loss": 0.3391, + "loss": 0.3593, "step": 636 }, { "epoch": 6.0, - "eval_LOCATION_f1": 0.5170068027210885, + "eval_LOCATION_f1": 0.4671532846715329, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.48717948717948717, - "eval_LOCATION_recall": 0.5507246376811594, - "eval_ORGANIZATION_f1": 0.3793103448275862, + "eval_LOCATION_precision": 0.47058823529411764, + "eval_LOCATION_recall": 0.463768115942029, + "eval_ORGANIZATION_f1": 0.39285714285714285, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.4, + "eval_ORGANIZATION_precision": 0.43137254901960786, "eval_ORGANIZATION_recall": 0.36065573770491804, - "eval_PERSON_f1": 0.7655786350148368, - "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.7010869565217391, - "eval_PERSON_recall": 0.8431372549019608, - "eval_QUANTITY_f1": 0.4262295081967213, - "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.4482758620689655, - "eval_QUANTITY_recall": 0.40625, - "eval_TIME_f1": 0.676470588235294, - "eval_TIME_number": 31, - "eval_TIME_precision": 0.6216216216216216, - "eval_TIME_recall": 0.7419354838709677, - "eval_loss": 0.2710684835910797, - "eval_overall_accuracy": 0.9251919742382958, - "eval_overall_f1": 0.617283950617284, - "eval_overall_precision": 0.587467362924282, - "eval_overall_recall": 0.6502890173410405, - "eval_runtime": 0.9395, - "eval_samples_per_second": 199.048, - "eval_steps_per_second": 3.193, + "eval_PERSON_f1": 0.711764705882353, + "eval_PERSON_number": 153, + "eval_PERSON_precision": 0.6470588235294118, + "eval_PERSON_recall": 0.7908496732026143, + "eval_QUANTITY_f1": 0.3389830508474576, + "eval_QUANTITY_number": 32, + "eval_QUANTITY_precision": 0.37037037037037035, + "eval_QUANTITY_recall": 0.3125, + "eval_TIME_f1": 0.6666666666666667, + "eval_TIME_number": 31, + "eval_TIME_precision": 0.65625, + "eval_TIME_recall": 0.6774193548387096, + "eval_loss": 0.2867273986339569, + "eval_overall_accuracy": 0.9227148872925439, + "eval_overall_f1": 0.5794655414908579, + "eval_overall_precision": 0.5643835616438356, + "eval_overall_recall": 0.5953757225433526, + "eval_runtime": 0.3704, + "eval_samples_per_second": 504.858, + "eval_steps_per_second": 8.099, "step": 636 }, { "epoch": 7.0, - "grad_norm": 0.7579200863838196, + "grad_norm": 0.9221069812774658, "learning_rate": 4.6500000000000005e-05, - "loss": 0.2822, + "loss": 0.296, "step": 742 }, { "epoch": 7.0, - "eval_LOCATION_f1": 0.6075949367088608, + "eval_LOCATION_f1": 0.5827814569536424, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.5393258426966292, - "eval_LOCATION_recall": 0.6956521739130435, - "eval_ORGANIZATION_f1": 0.5294117647058824, + "eval_LOCATION_precision": 0.5365853658536586, + "eval_LOCATION_recall": 0.6376811594202898, + "eval_ORGANIZATION_f1": 0.45925925925925926, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.48, - "eval_ORGANIZATION_recall": 0.5901639344262295, - "eval_PERSON_f1": 0.8263473053892217, + "eval_ORGANIZATION_precision": 0.4189189189189189, + "eval_ORGANIZATION_recall": 0.5081967213114754, + "eval_PERSON_f1": 0.8176470588235294, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.7624309392265194, - "eval_PERSON_recall": 0.9019607843137255, - "eval_QUANTITY_f1": 0.45714285714285713, + "eval_PERSON_precision": 0.7433155080213903, + "eval_PERSON_recall": 0.9084967320261438, + "eval_QUANTITY_f1": 0.463768115942029, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.42105263157894735, + "eval_QUANTITY_precision": 0.43243243243243246, "eval_QUANTITY_recall": 0.5, - "eval_TIME_f1": 0.7575757575757576, - "eval_TIME_number": 31, - "eval_TIME_precision": 0.7142857142857143, - "eval_TIME_recall": 0.8064516129032258, - "eval_loss": 0.23927439749240875, - "eval_overall_accuracy": 0.9343571959375774, - "eval_overall_f1": 0.6884816753926702, - "eval_overall_precision": 0.6291866028708134, - "eval_overall_recall": 0.7601156069364162, - "eval_runtime": 0.9433, - "eval_samples_per_second": 198.236, - "eval_steps_per_second": 3.18, + "eval_TIME_f1": 0.6567164179104478, + "eval_TIME_number": 31, + "eval_TIME_precision": 0.6111111111111112, + "eval_TIME_recall": 0.7096774193548387, + "eval_loss": 0.24547956883907318, + "eval_overall_accuracy": 0.9316324002972505, + "eval_overall_f1": 0.6614173228346456, + "eval_overall_precision": 0.6057692307692307, + "eval_overall_recall": 0.7283236994219653, + "eval_runtime": 0.3744, + "eval_samples_per_second": 499.493, + "eval_steps_per_second": 8.013, "step": 742 }, { "epoch": 8.0, - "grad_norm": 1.3045568466186523, + "grad_norm": 1.383445143699646, "learning_rate": 4.600000000000001e-05, - "loss": 0.2444, + "loss": 0.254, "step": 848 }, { "epoch": 8.0, - "eval_LOCATION_f1": 0.6794871794871794, + "eval_LOCATION_f1": 0.7006369426751592, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6091954022988506, - "eval_LOCATION_recall": 0.7681159420289855, - "eval_ORGANIZATION_f1": 0.554054054054054, + "eval_LOCATION_precision": 0.625, + "eval_LOCATION_recall": 0.7971014492753623, + "eval_ORGANIZATION_f1": 0.5100671140939597, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.47126436781609193, - "eval_ORGANIZATION_recall": 0.6721311475409836, - "eval_PERSON_f1": 0.8217522658610271, + "eval_ORGANIZATION_precision": 0.4318181818181818, + "eval_ORGANIZATION_recall": 0.6229508196721312, + "eval_PERSON_f1": 0.8249258160237388, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.7640449438202247, - "eval_PERSON_recall": 0.8888888888888888, + "eval_PERSON_precision": 0.7554347826086957, + "eval_PERSON_recall": 0.9084967320261438, "eval_QUANTITY_f1": 0.45714285714285713, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.42105263157894735, "eval_QUANTITY_recall": 0.5, - "eval_TIME_f1": 0.787878787878788, + "eval_TIME_f1": 0.7575757575757576, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7428571428571429, - "eval_TIME_recall": 0.8387096774193549, - "eval_loss": 0.20844174921512604, - "eval_overall_accuracy": 0.9383205350507803, - "eval_overall_f1": 0.7055771725032426, - "eval_overall_precision": 0.64, - "eval_overall_recall": 0.7861271676300579, - "eval_runtime": 0.9639, - "eval_samples_per_second": 194.01, - "eval_steps_per_second": 3.112, + "eval_TIME_precision": 0.7142857142857143, + "eval_TIME_recall": 0.8064516129032258, + "eval_loss": 0.2185853123664856, + "eval_overall_accuracy": 0.9375774089670548, + "eval_overall_f1": 0.7008985879332478, + "eval_overall_precision": 0.6304849884526559, + "eval_overall_recall": 0.7890173410404624, + "eval_runtime": 0.3742, + "eval_samples_per_second": 499.789, + "eval_steps_per_second": 8.018, "step": 848 }, { "epoch": 9.0, - "grad_norm": 1.5836372375488281, + "grad_norm": 1.6789393424987793, "learning_rate": 4.55e-05, - "loss": 0.2243, + "loss": 0.2277, "step": 954 }, { "epoch": 9.0, - "eval_LOCATION_f1": 0.7320261437908496, + "eval_LOCATION_f1": 0.7215189873417721, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6666666666666666, - "eval_LOCATION_recall": 0.8115942028985508, - "eval_ORGANIZATION_f1": 0.5915492957746479, + "eval_LOCATION_precision": 0.6404494382022472, + "eval_LOCATION_recall": 0.8260869565217391, + "eval_ORGANIZATION_f1": 0.5342465753424658, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5185185185185185, - "eval_ORGANIZATION_recall": 0.6885245901639344, - "eval_PERSON_f1": 0.8440366972477066, + "eval_ORGANIZATION_precision": 0.4588235294117647, + "eval_ORGANIZATION_recall": 0.639344262295082, + "eval_PERSON_f1": 0.8358208955223881, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.7931034482758621, - "eval_PERSON_recall": 0.9019607843137255, - "eval_QUANTITY_f1": 0.4999999999999999, + "eval_PERSON_precision": 0.7692307692307693, + "eval_PERSON_recall": 0.9150326797385621, + "eval_QUANTITY_f1": 0.5142857142857142, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.4722222222222222, - "eval_QUANTITY_recall": 0.53125, - "eval_TIME_f1": 0.787878787878788, + "eval_QUANTITY_precision": 0.47368421052631576, + "eval_QUANTITY_recall": 0.5625, + "eval_TIME_f1": 0.7812499999999999, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7428571428571429, - "eval_TIME_recall": 0.8387096774193549, - "eval_loss": 0.1914915144443512, - "eval_overall_accuracy": 0.9427792915531336, - "eval_overall_f1": 0.7380952380952382, - "eval_overall_precision": 0.6804878048780488, + "eval_TIME_precision": 0.7575757575757576, + "eval_TIME_recall": 0.8064516129032258, + "eval_loss": 0.1994231939315796, + "eval_overall_accuracy": 0.9403022046073817, + "eval_overall_f1": 0.7218628719275549, + "eval_overall_precision": 0.6533957845433255, "eval_overall_recall": 0.8063583815028902, - "eval_runtime": 0.9744, - "eval_samples_per_second": 191.917, - "eval_steps_per_second": 3.079, + "eval_runtime": 0.3766, + "eval_samples_per_second": 496.499, + "eval_steps_per_second": 7.965, "step": 954 }, { "epoch": 10.0, - "grad_norm": 1.011202096939087, + "grad_norm": 1.1687356233596802, "learning_rate": 4.5e-05, - "loss": 0.2035, + "loss": 0.2071, "step": 1060 }, { "epoch": 10.0, - "eval_LOCATION_f1": 0.763157894736842, + "eval_LOCATION_f1": 0.7712418300653594, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6987951807228916, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.5866666666666667, + "eval_LOCATION_precision": 0.7023809523809523, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.5430463576158939, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.4943820224719101, - "eval_ORGANIZATION_recall": 0.7213114754098361, - "eval_PERSON_f1": 0.851063829787234, + "eval_ORGANIZATION_precision": 0.45555555555555555, + "eval_ORGANIZATION_recall": 0.6721311475409836, + "eval_PERSON_f1": 0.8398791540785498, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.7954545454545454, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.5797101449275363, + "eval_PERSON_precision": 0.7808988764044944, + "eval_PERSON_recall": 0.9084967320261438, + "eval_QUANTITY_f1": 0.5507246376811593, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.5405405405405406, - "eval_QUANTITY_recall": 0.625, - "eval_TIME_f1": 0.7761194029850746, + "eval_QUANTITY_precision": 0.5135135135135135, + "eval_QUANTITY_recall": 0.59375, + "eval_TIME_f1": 0.7999999999999999, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7222222222222222, + "eval_TIME_precision": 0.7647058823529411, "eval_TIME_recall": 0.8387096774193549, - "eval_loss": 0.1796364188194275, - "eval_overall_accuracy": 0.9455040871934605, - "eval_overall_f1": 0.7509778357235984, - "eval_overall_precision": 0.684085510688836, - "eval_overall_recall": 0.8323699421965318, - "eval_runtime": 0.9675, - "eval_samples_per_second": 193.273, - "eval_steps_per_second": 3.101, + "eval_loss": 0.1809079945087433, + "eval_overall_accuracy": 0.9427792915531336, + "eval_overall_f1": 0.7386215864759428, + "eval_overall_precision": 0.6713947990543735, + "eval_overall_recall": 0.8208092485549133, + "eval_runtime": 0.3749, + "eval_samples_per_second": 498.839, + "eval_steps_per_second": 8.003, "step": 1060 }, { "epoch": 11.0, - "grad_norm": 1.90928316116333, + "grad_norm": 1.9759888648986816, "learning_rate": 4.4500000000000004e-05, - "loss": 0.1928, + "loss": 0.1937, "step": 1166 }, { "epoch": 11.0, - "eval_LOCATION_f1": 0.7388535031847134, + "eval_LOCATION_f1": 0.7692307692307693, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6590909090909091, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6013986013986014, + "eval_LOCATION_precision": 0.6896551724137931, + "eval_LOCATION_recall": 0.8695652173913043, + "eval_ORGANIZATION_f1": 0.5931034482758619, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.524390243902439, + "eval_ORGANIZATION_precision": 0.5119047619047619, "eval_ORGANIZATION_recall": 0.7049180327868853, - "eval_PERSON_f1": 0.8373493975903614, + "eval_PERSON_f1": 0.8459214501510575, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.776536312849162, - "eval_PERSON_recall": 0.9084967320261438, - "eval_QUANTITY_f1": 0.5797101449275363, + "eval_PERSON_precision": 0.7865168539325843, + "eval_PERSON_recall": 0.9150326797385621, + "eval_QUANTITY_f1": 0.6, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.5405405405405406, - "eval_QUANTITY_recall": 0.625, - "eval_TIME_f1": 0.7647058823529411, + "eval_QUANTITY_precision": 0.5526315789473685, + "eval_QUANTITY_recall": 0.65625, + "eval_TIME_f1": 0.7575757575757576, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7027027027027027, - "eval_TIME_recall": 0.8387096774193549, - "eval_loss": 0.16869711875915527, - "eval_overall_accuracy": 0.9469903393609116, - "eval_overall_f1": 0.7438231469440831, - "eval_overall_precision": 0.6761229314420804, - "eval_overall_recall": 0.8265895953757225, - "eval_runtime": 0.966, - "eval_samples_per_second": 193.581, - "eval_steps_per_second": 3.106, + "eval_TIME_precision": 0.7142857142857143, + "eval_TIME_recall": 0.8064516129032258, + "eval_loss": 0.17153412103652954, + "eval_overall_accuracy": 0.9455040871934605, + "eval_overall_f1": 0.7526041666666667, + "eval_overall_precision": 0.6848341232227488, + "eval_overall_recall": 0.8352601156069365, + "eval_runtime": 0.377, + "eval_samples_per_second": 496.013, + "eval_steps_per_second": 7.957, "step": 1166 }, { "epoch": 12.0, - "grad_norm": 2.2275421619415283, + "grad_norm": 2.159359931945801, "learning_rate": 4.4000000000000006e-05, - "loss": 0.1819, + "loss": 0.183, "step": 1272 }, { "epoch": 12.0, - "eval_LOCATION_f1": 0.7236842105263158, + "eval_LOCATION_f1": 0.7792207792207793, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6626506024096386, - "eval_LOCATION_recall": 0.7971014492753623, - "eval_ORGANIZATION_f1": 0.5942028985507246, + "eval_LOCATION_precision": 0.7058823529411765, + "eval_LOCATION_recall": 0.8695652173913043, + "eval_ORGANIZATION_f1": 0.5957446808510638, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5324675324675324, - "eval_ORGANIZATION_recall": 0.6721311475409836, - "eval_PERSON_f1": 0.8424242424242425, + "eval_ORGANIZATION_precision": 0.525, + "eval_ORGANIZATION_recall": 0.6885245901639344, + "eval_PERSON_f1": 0.8484848484848484, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.7853107344632768, - "eval_PERSON_recall": 0.9084967320261438, - "eval_QUANTITY_f1": 0.6376811594202898, + "eval_PERSON_precision": 0.7909604519774012, + "eval_PERSON_recall": 0.9150326797385621, + "eval_QUANTITY_f1": 0.6, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.5945945945945946, - "eval_QUANTITY_recall": 0.6875, - "eval_TIME_f1": 0.7647058823529411, + "eval_QUANTITY_precision": 0.5526315789473685, + "eval_QUANTITY_recall": 0.65625, + "eval_TIME_f1": 0.746268656716418, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7027027027027027, - "eval_TIME_recall": 0.8387096774193549, - "eval_loss": 0.1611860990524292, - "eval_overall_accuracy": 0.9464949219717612, - "eval_overall_f1": 0.7476882430647294, - "eval_overall_precision": 0.6885644768856448, - "eval_overall_recall": 0.8179190751445087, - "eval_runtime": 0.954, - "eval_samples_per_second": 196.012, - "eval_steps_per_second": 3.145, + "eval_TIME_precision": 0.6944444444444444, + "eval_TIME_recall": 0.8064516129032258, + "eval_loss": 0.16189619898796082, + "eval_overall_accuracy": 0.9479811741392122, + "eval_overall_f1": 0.7559055118110236, + "eval_overall_precision": 0.6923076923076923, + "eval_overall_recall": 0.8323699421965318, + "eval_runtime": 0.3748, + "eval_samples_per_second": 498.956, + "eval_steps_per_second": 8.005, "step": 1272 }, { "epoch": 13.0, - "grad_norm": 0.9414284825325012, + "grad_norm": 1.0158612728118896, "learning_rate": 4.35e-05, - "loss": 0.173, + "loss": 0.1739, "step": 1378 }, { "epoch": 13.0, - "eval_LOCATION_f1": 0.7770700636942676, + "eval_LOCATION_f1": 0.7843137254901961, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6931818181818182, - "eval_LOCATION_recall": 0.8840579710144928, - "eval_ORGANIZATION_f1": 0.5906040268456376, + "eval_LOCATION_precision": 0.7142857142857143, + "eval_LOCATION_recall": 0.8695652173913043, + "eval_ORGANIZATION_f1": 0.6068965517241379, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5, + "eval_ORGANIZATION_precision": 0.5238095238095238, "eval_ORGANIZATION_recall": 0.7213114754098361, - "eval_PERSON_f1": 0.8424242424242425, + "eval_PERSON_f1": 0.8571428571428571, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.7853107344632768, - "eval_PERSON_recall": 0.9084967320261438, - "eval_QUANTITY_f1": 0.6086956521739131, + "eval_PERSON_precision": 0.8011363636363636, + "eval_PERSON_recall": 0.9215686274509803, + "eval_QUANTITY_f1": 0.5797101449275363, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.5675675675675675, - "eval_QUANTITY_recall": 0.65625, - "eval_TIME_f1": 0.7761194029850746, + "eval_QUANTITY_precision": 0.5405405405405406, + "eval_QUANTITY_recall": 0.625, + "eval_TIME_f1": 0.746268656716418, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7222222222222222, - "eval_TIME_recall": 0.8387096774193549, - "eval_loss": 0.15597599744796753, - "eval_overall_accuracy": 0.9487243002229379, - "eval_overall_f1": 0.7538860103626942, - "eval_overall_precision": 0.6830985915492958, - "eval_overall_recall": 0.8410404624277457, - "eval_runtime": 0.9449, - "eval_samples_per_second": 197.9, - "eval_steps_per_second": 3.175, + "eval_TIME_precision": 0.6944444444444444, + "eval_TIME_recall": 0.8064516129032258, + "eval_loss": 0.15490159392356873, + "eval_overall_accuracy": 0.9512013871686896, + "eval_overall_f1": 0.7601572739187418, + "eval_overall_precision": 0.6954436450839329, + "eval_overall_recall": 0.838150289017341, + "eval_runtime": 0.372, + "eval_samples_per_second": 502.659, + "eval_steps_per_second": 8.064, "step": 1378 }, { "epoch": 14.0, - "grad_norm": 1.6199275255203247, + "grad_norm": 1.7819868326187134, "learning_rate": 4.3e-05, - "loss": 0.1716, + "loss": 0.1665, "step": 1484 }, { "epoch": 14.0, - "eval_LOCATION_f1": 0.7662337662337662, + "eval_LOCATION_f1": 0.7866666666666667, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6941176470588235, + "eval_LOCATION_precision": 0.7283950617283951, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.5962732919254659, + "eval_ORGANIZATION_f1": 0.6490066225165563, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.48, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.847560975609756, + "eval_ORGANIZATION_precision": 0.5444444444444444, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8685015290519879, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.7942857142857143, - "eval_PERSON_recall": 0.9084967320261438, - "eval_QUANTITY_f1": 0.6571428571428571, + "eval_PERSON_precision": 0.8160919540229885, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.626865671641791, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6052631578947368, - "eval_QUANTITY_recall": 0.71875, - "eval_TIME_f1": 0.7647058823529411, + "eval_QUANTITY_precision": 0.6, + "eval_QUANTITY_recall": 0.65625, + "eval_TIME_f1": 0.7692307692307693, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7027027027027027, - "eval_TIME_recall": 0.8387096774193549, - "eval_loss": 0.16146506369113922, - "eval_overall_accuracy": 0.9487243002229379, - "eval_overall_f1": 0.7554417413572343, - "eval_overall_precision": 0.6781609195402298, - "eval_overall_recall": 0.8526011560693642, - "eval_runtime": 0.9461, - "eval_samples_per_second": 197.651, - "eval_steps_per_second": 3.171, + "eval_TIME_precision": 0.7352941176470589, + "eval_TIME_recall": 0.8064516129032258, + "eval_loss": 0.15250568091869354, + "eval_overall_accuracy": 0.9529353480307159, + "eval_overall_f1": 0.7789473684210526, + "eval_overall_precision": 0.714975845410628, + "eval_overall_recall": 0.8554913294797688, + "eval_runtime": 0.3746, + "eval_samples_per_second": 499.251, + "eval_steps_per_second": 8.009, "step": 1484 }, { "epoch": 15.0, - "grad_norm": 0.7714660167694092, + "grad_norm": 0.672570526599884, "learning_rate": 4.25e-05, - "loss": 0.1614, + "loss": 0.1598, "step": 1590 }, { "epoch": 15.0, - "eval_LOCATION_f1": 0.7741935483870968, + "eval_LOCATION_f1": 0.7843137254901961, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6976744186046512, + "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6442953020134228, + "eval_ORGANIZATION_f1": 0.6206896551724138, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5454545454545454, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8313253012048193, + "eval_ORGANIZATION_precision": 0.5357142857142857, + "eval_ORGANIZATION_recall": 0.7377049180327869, + "eval_PERSON_f1": 0.8685015290519879, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.770949720670391, - "eval_PERSON_recall": 0.9019607843137255, - "eval_QUANTITY_f1": 0.5714285714285714, + "eval_PERSON_precision": 0.8160919540229885, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.6470588235294118, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.5263157894736842, - "eval_QUANTITY_recall": 0.625, - "eval_TIME_f1": 0.8059701492537312, + "eval_QUANTITY_precision": 0.6111111111111112, + "eval_QUANTITY_recall": 0.6875, + "eval_TIME_f1": 0.7647058823529411, "eval_TIME_number": 31, - "eval_TIME_precision": 0.75, - "eval_TIME_recall": 0.8709677419354839, - "eval_loss": 0.1512577384710312, - "eval_overall_accuracy": 0.9504582610849641, - "eval_overall_f1": 0.7580853816300128, - "eval_overall_precision": 0.6861826697892272, - "eval_overall_recall": 0.846820809248555, - "eval_runtime": 0.9698, - "eval_samples_per_second": 192.83, - "eval_steps_per_second": 3.094, + "eval_TIME_precision": 0.7027027027027027, + "eval_TIME_recall": 0.8387096774193549, + "eval_loss": 0.15064987540245056, + "eval_overall_accuracy": 0.9524399306415655, + "eval_overall_f1": 0.7752956636005256, + "eval_overall_precision": 0.7108433734939759, + "eval_overall_recall": 0.8526011560693642, + "eval_runtime": 0.3752, + "eval_samples_per_second": 498.368, + "eval_steps_per_second": 7.995, "step": 1590 }, { "epoch": 16.0, - "grad_norm": 4.456166744232178, + "grad_norm": 0.6825155019760132, "learning_rate": 4.2e-05, - "loss": 0.1595, + "loss": 0.1549, "step": 1696 }, { "epoch": 16.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7843137254901961, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.625, + "eval_LOCATION_precision": 0.7142857142857143, + "eval_LOCATION_recall": 0.8695652173913043, + "eval_ORGANIZATION_f1": 0.6388888888888888, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5421686746987951, - "eval_ORGANIZATION_recall": 0.7377049180327869, - "eval_PERSON_f1": 0.8404907975460123, + "eval_ORGANIZATION_precision": 0.5542168674698795, + "eval_ORGANIZATION_recall": 0.7540983606557377, + "eval_PERSON_f1": 0.8711656441717791, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.791907514450867, - "eval_PERSON_recall": 0.8954248366013072, - "eval_QUANTITY_f1": 0.5671641791044776, + "eval_PERSON_precision": 0.8208092485549133, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.626865671641791, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.5428571428571428, - "eval_QUANTITY_recall": 0.59375, - "eval_TIME_f1": 0.8181818181818182, + "eval_QUANTITY_precision": 0.6, + "eval_QUANTITY_recall": 0.65625, + "eval_TIME_f1": 0.782608695652174, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7714285714285715, + "eval_TIME_precision": 0.7105263157894737, "eval_TIME_recall": 0.8709677419354839, - "eval_loss": 0.1448289006948471, - "eval_overall_accuracy": 0.9526876393361406, - "eval_overall_f1": 0.7606382978723403, - "eval_overall_precision": 0.7044334975369458, - "eval_overall_recall": 0.8265895953757225, - "eval_runtime": 0.953, - "eval_samples_per_second": 196.23, - "eval_steps_per_second": 3.148, + "eval_loss": 0.1479189395904541, + "eval_overall_accuracy": 0.9536784741144414, + "eval_overall_f1": 0.7799736495388669, + "eval_overall_precision": 0.7167070217917676, + "eval_overall_recall": 0.8554913294797688, + "eval_runtime": 0.3703, + "eval_samples_per_second": 505.025, + "eval_steps_per_second": 8.102, "step": 1696 }, { "epoch": 17.0, - "grad_norm": 0.7077740430831909, + "grad_norm": 0.8551343083381653, "learning_rate": 4.15e-05, - "loss": 0.1525, + "loss": 0.1487, "step": 1802 }, { "epoch": 17.0, - "eval_LOCATION_f1": 0.7643312101910827, + "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6818181818181818, + "eval_LOCATION_precision": 0.6976744186046512, "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6225165562913907, + "eval_ORGANIZATION_f1": 0.64, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5222222222222223, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8440366972477066, + "eval_ORGANIZATION_precision": 0.5393258426966292, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8685015290519879, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.7931034482758621, - "eval_PERSON_recall": 0.9019607843137255, - "eval_QUANTITY_f1": 0.6388888888888888, + "eval_PERSON_precision": 0.8160919540229885, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.6857142857142857, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.575, - "eval_QUANTITY_recall": 0.71875, + "eval_QUANTITY_precision": 0.631578947368421, + "eval_QUANTITY_recall": 0.75, "eval_TIME_f1": 0.782608695652174, "eval_TIME_number": 31, "eval_TIME_precision": 0.7105263157894737, "eval_TIME_recall": 0.8709677419354839, - "eval_loss": 0.15009838342666626, - "eval_overall_accuracy": 0.9507059697795393, - "eval_overall_f1": 0.7603092783505155, - "eval_overall_precision": 0.686046511627907, - "eval_overall_recall": 0.8526011560693642, - "eval_runtime": 0.9578, - "eval_samples_per_second": 195.237, - "eval_steps_per_second": 3.132, + "eval_loss": 0.14963936805725098, + "eval_overall_accuracy": 0.9534307654198663, + "eval_overall_f1": 0.7808041504539559, + "eval_overall_precision": 0.7082352941176471, + "eval_overall_recall": 0.869942196531792, + "eval_runtime": 0.3724, + "eval_samples_per_second": 502.152, + "eval_steps_per_second": 8.056, "step": 1802 }, { "epoch": 18.0, - "grad_norm": 1.3755176067352295, + "grad_norm": 1.5059446096420288, "learning_rate": 4.1e-05, - "loss": 0.151, + "loss": 0.1471, "step": 1908 }, { "epoch": 18.0, - "eval_LOCATION_f1": 0.759493670886076, + "eval_LOCATION_f1": 0.7763157894736842, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6741573033707865, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6266666666666666, + "eval_LOCATION_precision": 0.7108433734939759, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6394557823129252, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5280898876404494, + "eval_ORGANIZATION_precision": 0.5465116279069767, "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8449848024316109, + "eval_PERSON_f1": 0.8711656441717791, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.7897727272727273, - "eval_PERSON_recall": 0.9084967320261438, - "eval_QUANTITY_f1": 0.6666666666666665, + "eval_PERSON_precision": 0.8208092485549133, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.6376811594202898, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6, - "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.823529411764706, + "eval_QUANTITY_precision": 0.5945945945945946, + "eval_QUANTITY_recall": 0.6875, + "eval_TIME_f1": 0.7428571428571428, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7567567567567568, - "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.14744549989700317, - "eval_overall_accuracy": 0.9524399306415655, - "eval_overall_f1": 0.7670527670527669, - "eval_overall_precision": 0.691415313225058, - "eval_overall_recall": 0.861271676300578, - "eval_runtime": 0.9533, - "eval_samples_per_second": 196.17, - "eval_steps_per_second": 3.147, + "eval_TIME_precision": 0.6666666666666666, + "eval_TIME_recall": 0.8387096774193549, + "eval_loss": 0.1425521969795227, + "eval_overall_accuracy": 0.9556601436710429, + "eval_overall_f1": 0.7748691099476441, + "eval_overall_precision": 0.7081339712918661, + "eval_overall_recall": 0.8554913294797688, + "eval_runtime": 0.3712, + "eval_samples_per_second": 503.811, + "eval_steps_per_second": 8.083, "step": 1908 }, { "epoch": 19.0, - "grad_norm": 4.061239719390869, + "grad_norm": 1.4319778680801392, "learning_rate": 4.05e-05, - "loss": 0.1457, + "loss": 0.1415, "step": 2014 }, { "epoch": 19.0, - "eval_LOCATION_f1": 0.7741935483870968, + "eval_LOCATION_f1": 0.8079470198675496, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6976744186046512, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6619718309859155, + "eval_LOCATION_precision": 0.7439024390243902, + "eval_LOCATION_recall": 0.8840579710144928, + "eval_ORGANIZATION_f1": 0.652482269503546, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5802469135802469, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8501529051987767, + "eval_ORGANIZATION_precision": 0.575, + "eval_ORGANIZATION_recall": 0.7540983606557377, + "eval_PERSON_f1": 0.8685015290519879, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.7988505747126436, - "eval_PERSON_recall": 0.9084967320261438, - "eval_QUANTITY_f1": 0.6478873239436619, + "eval_PERSON_precision": 0.8160919540229885, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.5897435897435898, - "eval_QUANTITY_recall": 0.71875, - "eval_TIME_f1": 0.8115942028985507, + "eval_QUANTITY_precision": 0.6756756756756757, + "eval_QUANTITY_recall": 0.78125, + "eval_TIME_f1": 0.7428571428571428, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7368421052631579, - "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.14025713503360748, - "eval_overall_accuracy": 0.9551647262818925, - "eval_overall_f1": 0.7774869109947645, - "eval_overall_precision": 0.7105263157894737, - "eval_overall_recall": 0.8583815028901735, - "eval_runtime": 0.9475, - "eval_samples_per_second": 197.359, - "eval_steps_per_second": 3.166, + "eval_TIME_precision": 0.6666666666666666, + "eval_TIME_recall": 0.8387096774193549, + "eval_loss": 0.1388433575630188, + "eval_overall_accuracy": 0.9568986871439188, + "eval_overall_f1": 0.7915567282321899, + "eval_overall_precision": 0.7281553398058253, + "eval_overall_recall": 0.8670520231213873, + "eval_runtime": 0.3716, + "eval_samples_per_second": 503.244, + "eval_steps_per_second": 8.073, "step": 2014 }, { "epoch": 20.0, - "grad_norm": 2.834820032119751, + "grad_norm": 0.6436260342597961, "learning_rate": 4e-05, - "loss": 0.1422, + "loss": 0.1385, "step": 2120 }, { "epoch": 20.0, - "eval_LOCATION_f1": 0.7721518987341773, + "eval_LOCATION_f1": 0.7870967741935485, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6853932584269663, + "eval_LOCATION_precision": 0.7093023255813954, "eval_LOCATION_recall": 0.8840579710144928, - "eval_ORGANIZATION_f1": 0.6225165562913907, + "eval_ORGANIZATION_f1": 0.6575342465753424, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5222222222222223, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.856269113149847, + "eval_ORGANIZATION_precision": 0.5647058823529412, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8606060606060606, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8045977011494253, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.6666666666666665, + "eval_PERSON_precision": 0.8022598870056498, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7536231884057971, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6, - "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.75, + "eval_QUANTITY_precision": 0.7027027027027027, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.8, "eval_TIME_number": 31, - "eval_TIME_precision": 0.6585365853658537, - "eval_TIME_recall": 0.8709677419354839, - "eval_loss": 0.1484678089618683, - "eval_overall_accuracy": 0.95169680455784, - "eval_overall_f1": 0.7666666666666667, - "eval_overall_precision": 0.6889400921658986, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 0.9645, - "eval_samples_per_second": 193.89, - "eval_steps_per_second": 3.111, + "eval_TIME_precision": 0.717948717948718, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1451817750930786, + "eval_overall_accuracy": 0.9546693088927422, + "eval_overall_f1": 0.7922077922077921, + "eval_overall_precision": 0.7193396226415094, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3763, + "eval_samples_per_second": 496.942, + "eval_steps_per_second": 7.972, "step": 2120 }, { "epoch": 21.0, - "grad_norm": 3.9201931953430176, + "grad_norm": 0.9739159941673279, "learning_rate": 3.9500000000000005e-05, - "loss": 0.1427, + "loss": 0.1361, "step": 2226 }, { "epoch": 21.0, - "eval_LOCATION_f1": 0.7843137254901961, + "eval_LOCATION_f1": 0.7712418300653594, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7142857142857143, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6216216216216216, + "eval_LOCATION_precision": 0.7023809523809523, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6442953020134228, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5287356321839081, - "eval_ORGANIZATION_recall": 0.7540983606557377, - "eval_PERSON_f1": 0.8527607361963191, + "eval_ORGANIZATION_precision": 0.5454545454545454, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8685015290519879, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8034682080924855, - "eval_PERSON_recall": 0.9084967320261438, - "eval_QUANTITY_f1": 0.6666666666666667, + "eval_PERSON_precision": 0.8160919540229885, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7164179104477612, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6216216216216216, - "eval_QUANTITY_recall": 0.71875, - "eval_TIME_f1": 0.8115942028985507, + "eval_QUANTITY_precision": 0.6857142857142857, + "eval_QUANTITY_recall": 0.75, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7368421052631579, + "eval_TIME_precision": 0.7567567567567568, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.14036431908607483, - "eval_overall_accuracy": 0.9549170175873173, - "eval_overall_f1": 0.7738562091503268, - "eval_overall_precision": 0.7064439140811456, - "eval_overall_recall": 0.8554913294797688, - "eval_runtime": 0.9596, - "eval_samples_per_second": 194.87, - "eval_steps_per_second": 3.126, + "eval_loss": 0.14020903408527374, + "eval_overall_accuracy": 0.955907852365618, + "eval_overall_f1": 0.7879581151832461, + "eval_overall_precision": 0.7200956937799043, + "eval_overall_recall": 0.869942196531792, + "eval_runtime": 0.3722, + "eval_samples_per_second": 502.459, + "eval_steps_per_second": 8.061, "step": 2226 }, { "epoch": 22.0, - "grad_norm": 0.7214634418487549, + "grad_norm": 1.1439884901046753, "learning_rate": 3.9000000000000006e-05, "loss": 0.1361, "step": 2332 }, { "epoch": 22.0, - "eval_LOCATION_f1": 0.7792207792207793, + "eval_LOCATION_f1": 0.7973856209150327, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7058823529411765, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.5986394557823129, + "eval_LOCATION_precision": 0.7261904761904762, + "eval_LOCATION_recall": 0.8840579710144928, + "eval_ORGANIZATION_f1": 0.64, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5116279069767442, - "eval_ORGANIZATION_recall": 0.7213114754098361, - "eval_PERSON_f1": 0.8597560975609756, + "eval_ORGANIZATION_precision": 0.5393258426966292, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8711656441717791, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8057142857142857, - "eval_PERSON_recall": 0.9215686274509803, - "eval_QUANTITY_f1": 0.676056338028169, + "eval_PERSON_precision": 0.8208092485549133, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7647058823529411, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6153846153846154, - "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.782608695652174, + "eval_QUANTITY_precision": 0.7222222222222222, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.8115942028985507, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7105263157894737, - "eval_TIME_recall": 0.8709677419354839, - "eval_loss": 0.13989298045635223, - "eval_overall_accuracy": 0.9539261828090166, - "eval_overall_f1": 0.7698309492847855, - "eval_overall_precision": 0.6997635933806147, - "eval_overall_recall": 0.8554913294797688, - "eval_runtime": 0.9637, - "eval_samples_per_second": 194.053, - "eval_steps_per_second": 3.113, + "eval_TIME_precision": 0.7368421052631579, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.14226193726062775, + "eval_overall_accuracy": 0.9549170175873173, + "eval_overall_f1": 0.7963446475195822, + "eval_overall_precision": 0.7261904761904762, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3724, + "eval_samples_per_second": 502.146, + "eval_steps_per_second": 8.056, "step": 2332 }, { "epoch": 23.0, - "grad_norm": 0.7087276577949524, + "grad_norm": 0.9229549765586853, "learning_rate": 3.85e-05, - "loss": 0.134, + "loss": 0.1331, "step": 2438 }, { "epoch": 23.0, - "eval_LOCATION_f1": 0.7894736842105263, + "eval_LOCATION_f1": 0.7947019867549668, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7228915662650602, + "eval_LOCATION_precision": 0.7317073170731707, "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.637037037037037, + "eval_ORGANIZATION_f1": 0.6474820143884893, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.581081081081081, - "eval_ORGANIZATION_recall": 0.7049180327868853, - "eval_PERSON_f1": 0.8492307692307692, + "eval_ORGANIZATION_precision": 0.5769230769230769, + "eval_ORGANIZATION_recall": 0.7377049180327869, + "eval_PERSON_f1": 0.8738461538461538, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8023255813953488, - "eval_PERSON_recall": 0.9019607843137255, - "eval_QUANTITY_f1": 0.6666666666666667, + "eval_PERSON_precision": 0.8255813953488372, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7462686567164178, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6216216216216216, - "eval_QUANTITY_recall": 0.71875, + "eval_QUANTITY_precision": 0.7142857142857143, + "eval_QUANTITY_recall": 0.78125, "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, "eval_TIME_precision": 0.7567567567567568, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.13254167139530182, - "eval_overall_accuracy": 0.9578895219222194, - "eval_overall_f1": 0.7797062750333779, - "eval_overall_precision": 0.7245657568238213, - "eval_overall_recall": 0.8439306358381503, - "eval_runtime": 0.9723, - "eval_samples_per_second": 192.324, - "eval_steps_per_second": 3.085, + "eval_loss": 0.13378849625587463, + "eval_overall_accuracy": 0.9581372306167947, + "eval_overall_f1": 0.7999999999999999, + "eval_overall_precision": 0.7425742574257426, + "eval_overall_recall": 0.8670520231213873, + "eval_runtime": 0.3737, + "eval_samples_per_second": 500.418, + "eval_steps_per_second": 8.028, "step": 2438 }, { "epoch": 24.0, - "grad_norm": 1.526930809020996, + "grad_norm": 1.7910873889923096, "learning_rate": 3.8e-05, - "loss": 0.132, + "loss": 0.1286, "step": 2544 }, { "epoch": 24.0, - "eval_LOCATION_f1": 0.7643312101910827, + "eval_LOCATION_f1": 0.7741935483870968, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.6818181818181818, + "eval_LOCATION_precision": 0.6976744186046512, "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6382978723404255, + "eval_ORGANIZATION_f1": 0.6621621621621622, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5625, - "eval_ORGANIZATION_recall": 0.7377049180327869, - "eval_PERSON_f1": 0.8623853211009175, + "eval_ORGANIZATION_precision": 0.5632183908045977, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8738461538461538, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8103448275862069, - "eval_PERSON_recall": 0.9215686274509803, - "eval_QUANTITY_f1": 0.6944444444444444, + "eval_PERSON_precision": 0.8255813953488372, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7222222222222223, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.625, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.7714285714285716, + "eval_QUANTITY_precision": 0.65, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.7887323943661972, "eval_TIME_number": 31, - "eval_TIME_precision": 0.6923076923076923, - "eval_TIME_recall": 0.8709677419354839, - "eval_loss": 0.14164899289608002, - "eval_overall_accuracy": 0.9546693088927422, - "eval_overall_f1": 0.7770534550195568, - "eval_overall_precision": 0.7078384798099763, - "eval_overall_recall": 0.861271676300578, - "eval_runtime": 0.9598, - "eval_samples_per_second": 194.835, - "eval_steps_per_second": 3.126, + "eval_TIME_precision": 0.7, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1469903588294983, + "eval_overall_accuracy": 0.9544216001981669, + "eval_overall_f1": 0.7911802853437094, + "eval_overall_precision": 0.7176470588235294, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3719, + "eval_samples_per_second": 502.77, + "eval_steps_per_second": 8.066, "step": 2544 }, { "epoch": 25.0, - "grad_norm": 1.102123498916626, + "grad_norm": 1.3032689094543457, "learning_rate": 3.7500000000000003e-05, - "loss": 0.1305, + "loss": 0.1274, "step": 2650 }, { "epoch": 25.0, - "eval_LOCATION_f1": 0.7792207792207793, + "eval_LOCATION_f1": 0.7843137254901961, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7058823529411765, + "eval_LOCATION_precision": 0.7142857142857143, "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6344827586206897, + "eval_ORGANIZATION_f1": 0.6620689655172414, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5476190476190477, - "eval_ORGANIZATION_recall": 0.7540983606557377, - "eval_PERSON_f1": 0.8571428571428571, + "eval_ORGANIZATION_precision": 0.5714285714285714, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8676923076923077, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8011363636363636, + "eval_PERSON_precision": 0.8197674418604651, "eval_PERSON_recall": 0.9215686274509803, - "eval_QUANTITY_f1": 0.6944444444444444, + "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.625, + "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.7714285714285716, + "eval_TIME_f1": 0.8115942028985507, "eval_TIME_number": 31, - "eval_TIME_precision": 0.6923076923076923, - "eval_TIME_recall": 0.8709677419354839, - "eval_loss": 0.13924439251422882, - "eval_overall_accuracy": 0.9544216001981669, - "eval_overall_f1": 0.7766233766233767, - "eval_overall_precision": 0.7051886792452831, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 0.9576, - "eval_samples_per_second": 195.277, - "eval_steps_per_second": 3.133, + "eval_TIME_precision": 0.7368421052631579, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.13992294669151306, + "eval_overall_accuracy": 0.9566509784493435, + "eval_overall_f1": 0.7936925098554534, + "eval_overall_precision": 0.727710843373494, + "eval_overall_recall": 0.8728323699421965, + "eval_runtime": 0.3712, + "eval_samples_per_second": 503.751, + "eval_steps_per_second": 8.082, "step": 2650 }, { "epoch": 26.0, - "grad_norm": 0.8809081315994263, + "grad_norm": 1.104667067527771, "learning_rate": 3.7e-05, - "loss": 0.1278, + "loss": 0.1269, "step": 2756 }, { "epoch": 26.0, - "eval_LOCATION_f1": 0.7894736842105263, + "eval_LOCATION_f1": 0.8026315789473685, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7228915662650602, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6266666666666666, + "eval_LOCATION_precision": 0.7349397590361446, + "eval_LOCATION_recall": 0.8840579710144928, + "eval_ORGANIZATION_f1": 0.6490066225165563, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5280898876404494, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8623853211009175, + "eval_ORGANIZATION_precision": 0.5444444444444444, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8685015290519879, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8103448275862069, - "eval_PERSON_recall": 0.9215686274509803, - "eval_QUANTITY_f1": 0.7042253521126761, + "eval_PERSON_precision": 0.8160919540229885, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7647058823529411, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6410256410256411, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8, + "eval_QUANTITY_precision": 0.7222222222222222, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.717948717948718, + "eval_TIME_precision": 0.7567567567567568, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.13592180609703064, - "eval_overall_accuracy": 0.9554124349764677, - "eval_overall_f1": 0.781818181818182, - "eval_overall_precision": 0.7099056603773585, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9573, - "eval_samples_per_second": 195.338, - "eval_steps_per_second": 3.134, + "eval_loss": 0.14159581065177917, + "eval_overall_accuracy": 0.9556601436710429, + "eval_overall_f1": 0.7989556135770234, + "eval_overall_precision": 0.7285714285714285, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.3741, + "eval_samples_per_second": 499.897, + "eval_steps_per_second": 8.02, "step": 2756 }, { "epoch": 27.0, - "grad_norm": 1.6803548336029053, + "grad_norm": 3.7249088287353516, "learning_rate": 3.65e-05, - "loss": 0.1257, + "loss": 0.1242, "step": 2862 }, { "epoch": 27.0, - "eval_LOCATION_f1": 0.7973856209150327, + "eval_LOCATION_f1": 0.7763157894736842, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7261904761904762, - "eval_LOCATION_recall": 0.8840579710144928, - "eval_ORGANIZATION_f1": 0.6474820143884893, + "eval_LOCATION_precision": 0.7108433734939759, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6857142857142857, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5769230769230769, - "eval_ORGANIZATION_recall": 0.7377049180327869, - "eval_PERSON_f1": 0.8738461538461538, + "eval_ORGANIZATION_precision": 0.6075949367088608, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8685015290519879, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8255813953488372, + "eval_PERSON_precision": 0.8160919540229885, "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.6944444444444444, + "eval_QUANTITY_f1": 0.7536231884057971, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.625, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8, + "eval_QUANTITY_precision": 0.7027027027027027, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.717948717948718, + "eval_TIME_precision": 0.7567567567567568, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.13588088750839233, - "eval_overall_accuracy": 0.9564032697547684, - "eval_overall_f1": 0.7931488801054017, - "eval_overall_precision": 0.7288135593220338, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9579, - "eval_samples_per_second": 195.215, - "eval_steps_per_second": 3.132, + "eval_loss": 0.13718503713607788, + "eval_overall_accuracy": 0.9568986871439188, + "eval_overall_f1": 0.8015873015873016, + "eval_overall_precision": 0.7390243902439024, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3743, + "eval_samples_per_second": 499.541, + "eval_steps_per_second": 8.014, "step": 2862 }, { "epoch": 28.0, - "grad_norm": 1.4917080402374268, + "grad_norm": 4.685704231262207, "learning_rate": 3.6e-05, - "loss": 0.1249, + "loss": 0.1223, "step": 2968 }, { "epoch": 28.0, - "eval_LOCATION_f1": 0.7947019867549668, + "eval_LOCATION_f1": 0.8187919463087248, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7317073170731707, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6428571428571429, + "eval_LOCATION_precision": 0.7625, + "eval_LOCATION_recall": 0.8840579710144928, + "eval_ORGANIZATION_f1": 0.6619718309859155, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.569620253164557, - "eval_ORGANIZATION_recall": 0.7377049180327869, - "eval_PERSON_f1": 0.8650306748466258, + "eval_ORGANIZATION_precision": 0.5802469135802469, + "eval_ORGANIZATION_recall": 0.7704918032786885, + "eval_PERSON_f1": 0.8738461538461538, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.815028901734104, - "eval_PERSON_recall": 0.9215686274509803, - "eval_QUANTITY_f1": 0.6857142857142857, + "eval_PERSON_precision": 0.8255813953488372, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7647058823529411, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.631578947368421, - "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.8529411764705881, + "eval_QUANTITY_precision": 0.7222222222222222, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1313154250383377, - "eval_overall_accuracy": 0.9583849393113698, - "eval_overall_f1": 0.7920529801324503, - "eval_overall_precision": 0.7310513447432763, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 1.0156, - "eval_samples_per_second": 184.125, - "eval_steps_per_second": 2.954, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1325845867395401, + "eval_overall_accuracy": 0.9588803567005202, + "eval_overall_f1": 0.8095872170439414, + "eval_overall_precision": 0.7506172839506173, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3723, + "eval_samples_per_second": 502.217, + "eval_steps_per_second": 8.057, "step": 2968 }, { "epoch": 29.0, - "grad_norm": 1.344867467880249, + "grad_norm": 3.4481709003448486, "learning_rate": 3.55e-05, - "loss": 0.1233, + "loss": 0.1193, "step": 3074 }, { "epoch": 29.0, - "eval_LOCATION_f1": 0.8026315789473685, + "eval_LOCATION_f1": 0.7999999999999999, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7349397590361446, - "eval_LOCATION_recall": 0.8840579710144928, - "eval_ORGANIZATION_f1": 0.6142857142857143, + "eval_LOCATION_precision": 0.7407407407407407, + "eval_LOCATION_recall": 0.8695652173913043, + "eval_ORGANIZATION_f1": 0.6762589928057554, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5443037974683544, - "eval_ORGANIZATION_recall": 0.7049180327868853, - "eval_PERSON_f1": 0.8703703703703703, + "eval_ORGANIZATION_precision": 0.6025641025641025, + "eval_ORGANIZATION_recall": 0.7704918032786885, + "eval_PERSON_f1": 0.8819875776397514, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8245614035087719, - "eval_PERSON_recall": 0.9215686274509803, - "eval_QUANTITY_f1": 0.676470588235294, + "eval_PERSON_precision": 0.8402366863905325, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7647058823529411, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6388888888888888, - "eval_QUANTITY_recall": 0.71875, - "eval_TIME_f1": 0.835820895522388, + "eval_QUANTITY_precision": 0.7222222222222222, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_precision": 0.7567567567567568, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.13240520656108856, - "eval_overall_accuracy": 0.9573941045330691, - "eval_overall_f1": 0.7882822902796272, - "eval_overall_precision": 0.7308641975308642, - "eval_overall_recall": 0.8554913294797688, - "eval_runtime": 0.9601, - "eval_samples_per_second": 194.771, - "eval_steps_per_second": 3.125, + "eval_loss": 0.1309487521648407, + "eval_overall_accuracy": 0.9596234827842457, + "eval_overall_f1": 0.8112449799196787, + "eval_overall_precision": 0.7556109725685786, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3737, + "eval_samples_per_second": 500.387, + "eval_steps_per_second": 8.028, "step": 3074 }, { "epoch": 30.0, - "grad_norm": 0.9224383234977722, + "grad_norm": 0.7676149606704712, "learning_rate": 3.5e-05, - "loss": 0.1209, + "loss": 0.119, "step": 3180 }, { "epoch": 30.0, - "eval_LOCATION_f1": 0.7763157894736842, + "eval_LOCATION_f1": 0.8053691275167786, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7108433734939759, - "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6577181208053691, + "eval_LOCATION_precision": 0.75, + "eval_LOCATION_recall": 0.8695652173913043, + "eval_ORGANIZATION_f1": 0.6666666666666666, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5568181818181818, - "eval_ORGANIZATION_recall": 0.8032786885245902, - "eval_PERSON_f1": 0.8703703703703703, + "eval_ORGANIZATION_precision": 0.5875, + "eval_ORGANIZATION_recall": 0.7704918032786885, + "eval_PERSON_f1": 0.8792569659442725, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8245614035087719, - "eval_PERSON_recall": 0.9215686274509803, - "eval_QUANTITY_f1": 0.676056338028169, + "eval_PERSON_precision": 0.8352941176470589, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7536231884057971, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6153846153846154, - "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.8115942028985507, + "eval_QUANTITY_precision": 0.7027027027027027, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7368421052631579, + "eval_TIME_precision": 0.7777777777777778, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.13591422140598297, - "eval_overall_accuracy": 0.9561555610601932, - "eval_overall_f1": 0.7869281045751635, - "eval_overall_precision": 0.7183770883054893, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9514, - "eval_samples_per_second": 196.556, - "eval_steps_per_second": 3.153, + "eval_loss": 0.13215813040733337, + "eval_overall_accuracy": 0.9601189001733961, + "eval_overall_f1": 0.8090787716955942, + "eval_overall_precision": 0.7518610421836228, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3736, + "eval_samples_per_second": 500.504, + "eval_steps_per_second": 8.029, "step": 3180 }, { "epoch": 31.0, - "grad_norm": 0.6330462098121643, + "grad_norm": 0.5464662313461304, "learning_rate": 3.45e-05, - "loss": 0.1196, + "loss": 0.1208, "step": 3286 }, { "epoch": 31.0, - "eval_LOCATION_f1": 0.7894736842105263, + "eval_LOCATION_f1": 0.7866666666666667, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7228915662650602, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6433566433566433, + "eval_LOCATION_precision": 0.7283950617283951, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6713286713286712, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5609756097560976, - "eval_ORGANIZATION_recall": 0.7540983606557377, - "eval_PERSON_f1": 0.8685015290519879, + "eval_ORGANIZATION_precision": 0.5853658536585366, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8765432098765433, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8160919540229885, + "eval_PERSON_precision": 0.8304093567251462, "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.6857142857142857, + "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.631578947368421, - "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.8285714285714285, + "eval_QUANTITY_precision": 0.6756756756756757, + "eval_QUANTITY_recall": 0.78125, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7435897435897436, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.13350899517536163, - "eval_overall_accuracy": 0.9566509784493435, - "eval_overall_f1": 0.7900262467191601, - "eval_overall_precision": 0.7235576923076923, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.956, - "eval_samples_per_second": 195.61, - "eval_steps_per_second": 3.138, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.13540008664131165, + "eval_overall_accuracy": 0.958632648005945, + "eval_overall_f1": 0.8010610079575597, + "eval_overall_precision": 0.7401960784313726, + "eval_overall_recall": 0.8728323699421965, + "eval_runtime": 0.3719, + "eval_samples_per_second": 502.807, + "eval_steps_per_second": 8.066, "step": 3286 }, { "epoch": 32.0, - "grad_norm": 0.9359336495399475, + "grad_norm": 1.3645358085632324, "learning_rate": 3.4000000000000007e-05, - "loss": 0.1195, + "loss": 0.1143, "step": 3392 }, { "epoch": 32.0, - "eval_LOCATION_f1": 0.7843137254901961, + "eval_LOCATION_f1": 0.7866666666666667, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7142857142857143, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6187050359712231, + "eval_LOCATION_precision": 0.7283950617283951, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6573426573426574, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5512820512820513, - "eval_ORGANIZATION_recall": 0.7049180327868853, - "eval_PERSON_f1": 0.8765432098765433, + "eval_ORGANIZATION_precision": 0.573170731707317, + "eval_ORGANIZATION_recall": 0.7704918032786885, + "eval_PERSON_f1": 0.8792569659442725, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8304093567251462, + "eval_PERSON_precision": 0.8352941176470589, "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7058823529411765, + "eval_QUANTITY_f1": 0.7462686567164178, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6666666666666666, - "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.8115942028985507, + "eval_QUANTITY_precision": 0.7142857142857143, + "eval_QUANTITY_recall": 0.78125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7368421052631579, + "eval_TIME_precision": 0.7777777777777778, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.13091404736042023, - "eval_overall_accuracy": 0.9578895219222194, - "eval_overall_f1": 0.7888446215139442, - "eval_overall_precision": 0.7297297297297297, - "eval_overall_recall": 0.8583815028901735, - "eval_runtime": 0.9602, - "eval_samples_per_second": 194.755, - "eval_steps_per_second": 3.124, + "eval_loss": 0.13142676651477814, + "eval_overall_accuracy": 0.958632648005945, + "eval_overall_f1": 0.8026666666666668, + "eval_overall_precision": 0.745049504950495, + "eval_overall_recall": 0.869942196531792, + "eval_runtime": 0.3729, + "eval_samples_per_second": 501.416, + "eval_steps_per_second": 8.044, "step": 3392 }, { "epoch": 33.0, - "grad_norm": 2.8279988765716553, + "grad_norm": 1.6423691511154175, "learning_rate": 3.35e-05, - "loss": 0.1183, + "loss": 0.1151, "step": 3498 }, { @@ -1269,232 +1269,232 @@ "eval_LOCATION_number": 69, "eval_LOCATION_precision": 0.7317073170731707, "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.647887323943662, + "eval_ORGANIZATION_f1": 0.6620689655172414, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5679012345679012, - "eval_ORGANIZATION_recall": 0.7540983606557377, - "eval_PERSON_f1": 0.8792569659442725, + "eval_ORGANIZATION_precision": 0.5714285714285714, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8765432098765433, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8352941176470589, + "eval_PERSON_precision": 0.8304093567251462, "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7164179104477612, + "eval_QUANTITY_f1": 0.6956521739130435, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6857142857142857, + "eval_QUANTITY_precision": 0.6486486486486487, "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.8115942028985507, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7368421052631579, + "eval_TIME_precision": 0.7567567567567568, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.13136102259159088, - "eval_overall_accuracy": 0.958632648005945, - "eval_overall_f1": 0.7978723404255319, - "eval_overall_precision": 0.7389162561576355, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 1.005, - "eval_samples_per_second": 186.067, - "eval_steps_per_second": 2.985, + "eval_loss": 0.132831409573555, + "eval_overall_accuracy": 0.9591280653950953, + "eval_overall_f1": 0.797886393659181, + "eval_overall_precision": 0.7347931873479319, + "eval_overall_recall": 0.8728323699421965, + "eval_runtime": 0.3744, + "eval_samples_per_second": 499.492, + "eval_steps_per_second": 8.013, "step": 3498 }, { "epoch": 34.0, - "grad_norm": 0.6278368830680847, + "grad_norm": 0.8335278630256653, "learning_rate": 3.3e-05, - "loss": 0.1142, + "loss": 0.1121, "step": 3604 }, { "epoch": 34.0, - "eval_LOCATION_f1": 0.7712418300653594, + "eval_LOCATION_f1": 0.7999999999999999, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7023809523809523, - "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6405228758169935, + "eval_LOCATION_precision": 0.7407407407407407, + "eval_LOCATION_recall": 0.8695652173913043, + "eval_ORGANIZATION_f1": 0.6533333333333333, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.532608695652174, + "eval_ORGANIZATION_precision": 0.550561797752809, "eval_ORGANIZATION_recall": 0.8032786885245902, - "eval_PERSON_f1": 0.8792569659442725, + "eval_PERSON_f1": 0.8800000000000001, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8352941176470589, - "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.676056338028169, + "eval_PERSON_precision": 0.8313953488372093, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.7536231884057971, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6153846153846154, - "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.8115942028985507, - "eval_TIME_number": 31, - "eval_TIME_precision": 0.7368421052631579, + "eval_QUANTITY_precision": 0.7027027027027027, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.823529411764706, + "eval_TIME_number": 31, + "eval_TIME_precision": 0.7567567567567568, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.13677182793617249, - "eval_overall_accuracy": 0.9546693088927422, - "eval_overall_f1": 0.7854356306892067, - "eval_overall_precision": 0.7139479905437353, - "eval_overall_recall": 0.8728323699421965, - "eval_runtime": 0.9562, - "eval_samples_per_second": 195.567, - "eval_steps_per_second": 3.137, + "eval_loss": 0.13941965997219086, + "eval_overall_accuracy": 0.9571463958384939, + "eval_overall_f1": 0.8031496062992126, + "eval_overall_precision": 0.7355769230769231, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.375, + "eval_samples_per_second": 498.612, + "eval_steps_per_second": 7.999, "step": 3604 }, { "epoch": 35.0, - "grad_norm": 2.2821428775787354, + "grad_norm": 2.70279598236084, "learning_rate": 3.2500000000000004e-05, - "loss": 0.1175, + "loss": 0.1149, "step": 3710 }, { "epoch": 35.0, - "eval_LOCATION_f1": 0.7894736842105263, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7228915662650602, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6619718309859155, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6486486486486487, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5802469135802469, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8792569659442725, + "eval_ORGANIZATION_precision": 0.5517241379310345, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8827160493827161, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8352941176470589, - "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.6956521739130435, + "eval_PERSON_precision": 0.8362573099415205, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.7536231884057971, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6486486486486487, - "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.8405797101449276, + "eval_QUANTITY_precision": 0.7027027027027027, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1275167614221573, - "eval_overall_accuracy": 0.9588803567005202, - "eval_overall_f1": 0.7999999999999999, - "eval_overall_precision": 0.7383863080684596, - "eval_overall_recall": 0.8728323699421965, - "eval_runtime": 0.9572, - "eval_samples_per_second": 195.368, - "eval_steps_per_second": 3.134, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.13413338363170624, + "eval_overall_accuracy": 0.9583849393113698, + "eval_overall_f1": 0.8031704095112285, + "eval_overall_precision": 0.7396593673965937, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3724, + "eval_samples_per_second": 502.185, + "eval_steps_per_second": 8.056, "step": 3710 }, { "epoch": 36.0, - "grad_norm": 1.224282145500183, + "grad_norm": 0.7904542684555054, "learning_rate": 3.2000000000000005e-05, - "loss": 0.1187, + "loss": 0.1114, "step": 3816 }, { "epoch": 36.0, - "eval_LOCATION_f1": 0.7947019867549668, + "eval_LOCATION_f1": 0.7972972972972974, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7317073170731707, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6521739130434783, + "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6857142857142857, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5844155844155844, - "eval_ORGANIZATION_recall": 0.7377049180327869, - "eval_PERSON_f1": 0.8792569659442725, + "eval_ORGANIZATION_precision": 0.6075949367088608, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8854489164086689, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8352941176470589, - "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7042253521126761, + "eval_PERSON_precision": 0.8411764705882353, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6410256410256411, + "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.823529411764706, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_precision": 0.7777777777777778, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.1280500590801239, - "eval_overall_accuracy": 0.9598711914788209, - "eval_overall_f1": 0.7989347536617841, - "eval_overall_precision": 0.7407407407407407, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9495, - "eval_samples_per_second": 196.942, - "eval_steps_per_second": 3.16, + "eval_loss": 0.12858277559280396, + "eval_overall_accuracy": 0.9606143175625464, + "eval_overall_f1": 0.8112449799196787, + "eval_overall_precision": 0.7556109725685786, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3711, + "eval_samples_per_second": 503.902, + "eval_steps_per_second": 8.084, "step": 3816 }, { "epoch": 37.0, - "grad_norm": 1.6785736083984375, + "grad_norm": 1.3650281429290771, "learning_rate": 3.15e-05, - "loss": 0.1137, + "loss": 0.1092, "step": 3922 }, { "epoch": 37.0, - "eval_LOCATION_f1": 0.8133333333333332, + "eval_LOCATION_f1": 0.7866666666666667, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7530864197530864, - "eval_LOCATION_recall": 0.8840579710144928, - "eval_ORGANIZATION_f1": 0.6580645161290322, + "eval_LOCATION_precision": 0.7283950617283951, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6805555555555556, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5425531914893617, - "eval_ORGANIZATION_recall": 0.8360655737704918, - "eval_PERSON_f1": 0.8765432098765433, + "eval_ORGANIZATION_precision": 0.5903614457831325, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8800000000000001, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8304093567251462, - "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7042253521126761, + "eval_PERSON_precision": 0.8313953488372093, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6410256410256411, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8115942028985507, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7368421052631579, + "eval_TIME_precision": 0.7777777777777778, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.13252070546150208, - "eval_overall_accuracy": 0.9564032697547684, - "eval_overall_f1": 0.7984395318595578, - "eval_overall_precision": 0.7257683215130024, - "eval_overall_recall": 0.8872832369942196, - "eval_runtime": 0.9687, - "eval_samples_per_second": 193.043, - "eval_steps_per_second": 3.097, + "eval_loss": 0.13275323808193207, + "eval_overall_accuracy": 0.9593757740896706, + "eval_overall_f1": 0.8068783068783069, + "eval_overall_precision": 0.7439024390243902, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3737, + "eval_samples_per_second": 500.384, + "eval_steps_per_second": 8.028, "step": 3922 }, { "epoch": 38.0, - "grad_norm": 1.6354094743728638, + "grad_norm": 1.638619065284729, "learning_rate": 3.1e-05, - "loss": 0.1131, + "loss": 0.1072, "step": 4028 }, { "epoch": 38.0, - "eval_LOCATION_f1": 0.8108108108108109, + "eval_LOCATION_f1": 0.7866666666666667, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.759493670886076, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6329113924050632, + "eval_LOCATION_precision": 0.7283950617283951, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6853146853146853, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5154639175257731, - "eval_ORGANIZATION_recall": 0.819672131147541, - "eval_PERSON_f1": 0.8658536585365854, + "eval_ORGANIZATION_precision": 0.5975609756097561, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8773006134969324, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8114285714285714, - "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7042253521126761, + "eval_PERSON_precision": 0.8265895953757225, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6410256410256411, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.717948717948718, + "eval_TIME_precision": 0.7777777777777778, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.1375594586133957, - "eval_overall_accuracy": 0.9536784741144414, - "eval_overall_f1": 0.7870967741935484, - "eval_overall_precision": 0.710955710955711, + "eval_loss": 0.133976012468338, + "eval_overall_accuracy": 0.9598711914788209, + "eval_overall_f1": 0.8068783068783069, + "eval_overall_precision": 0.7439024390243902, "eval_overall_recall": 0.8815028901734104, - "eval_runtime": 0.9604, - "eval_samples_per_second": 194.719, - "eval_steps_per_second": 3.124, + "eval_runtime": 0.3719, + "eval_samples_per_second": 502.881, + "eval_steps_per_second": 8.068, "step": 4028 }, { "epoch": 39.0, - "grad_norm": 0.7580879926681519, + "grad_norm": 0.7919366955757141, "learning_rate": 3.05e-05, - "loss": 0.1112, + "loss": 0.1032, "step": 4134 }, { @@ -1503,544 +1503,544 @@ "eval_LOCATION_number": 69, "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6376811594202898, + "eval_ORGANIZATION_f1": 0.6713286713286712, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5714285714285714, - "eval_ORGANIZATION_recall": 0.7213114754098361, - "eval_PERSON_f1": 0.8685015290519879, + "eval_ORGANIZATION_precision": 0.5853658536585366, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8765432098765433, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8160919540229885, + "eval_PERSON_precision": 0.8304093567251462, "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7352941176470588, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6944444444444444, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1263578087091446, - "eval_overall_accuracy": 0.958632648005945, - "eval_overall_f1": 0.796271637816245, - "eval_overall_precision": 0.7382716049382716, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 0.9588, - "eval_samples_per_second": 195.03, - "eval_steps_per_second": 3.129, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1285211592912674, + "eval_overall_accuracy": 0.9603666088679713, + "eval_overall_f1": 0.804780876494024, + "eval_overall_precision": 0.7444717444717445, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3722, + "eval_samples_per_second": 502.429, + "eval_steps_per_second": 8.06, "step": 4134 }, { "epoch": 40.0, - "grad_norm": 1.0427302122116089, + "grad_norm": 1.121160626411438, "learning_rate": 3e-05, - "loss": 0.1076, + "loss": 0.1045, "step": 4240 }, { "epoch": 40.0, - "eval_LOCATION_f1": 0.7763157894736842, + "eval_LOCATION_f1": 0.7866666666666667, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7108433734939759, + "eval_LOCATION_precision": 0.7283950617283951, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.647887323943662, + "eval_ORGANIZATION_f1": 0.676056338028169, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5679012345679012, - "eval_ORGANIZATION_recall": 0.7540983606557377, - "eval_PERSON_f1": 0.8711656441717791, + "eval_ORGANIZATION_precision": 0.5925925925925926, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8765432098765433, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8208092485549133, + "eval_PERSON_precision": 0.8304093567251462, "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7042253521126761, + "eval_QUANTITY_f1": 0.7536231884057971, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6410256410256411, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, - "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12935374677181244, - "eval_overall_accuracy": 0.9578895219222194, - "eval_overall_f1": 0.7921052631578948, - "eval_overall_precision": 0.7270531400966184, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9564, - "eval_samples_per_second": 195.529, - "eval_steps_per_second": 3.137, + "eval_QUANTITY_precision": 0.7027027027027027, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, + "eval_TIME_number": 31, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12849295139312744, + "eval_overall_accuracy": 0.961357443646272, + "eval_overall_f1": 0.8058510638297872, + "eval_overall_precision": 0.7463054187192119, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3741, + "eval_samples_per_second": 499.806, + "eval_steps_per_second": 8.018, "step": 4240 }, { "epoch": 41.0, - "grad_norm": 1.631004810333252, + "grad_norm": 1.5202959775924683, "learning_rate": 2.95e-05, - "loss": 0.1085, + "loss": 0.1057, "step": 4346 }, { "epoch": 41.0, - "eval_LOCATION_f1": 0.7814569536423841, + "eval_LOCATION_f1": 0.7866666666666667, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7195121951219512, + "eval_LOCATION_precision": 0.7283950617283951, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6438356164383562, + "eval_ORGANIZATION_f1": 0.6944444444444445, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5529411764705883, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8792569659442725, + "eval_ORGANIZATION_precision": 0.6024096385542169, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8773006134969324, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8352941176470589, - "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7142857142857143, + "eval_PERSON_precision": 0.8265895953757225, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6578947368421053, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8285714285714285, - "eval_TIME_number": 31, - "eval_TIME_precision": 0.7435897435897436, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1293003410100937, - "eval_overall_accuracy": 0.9578895219222194, - "eval_overall_f1": 0.7947368421052632, - "eval_overall_precision": 0.7294685990338164, - "eval_overall_recall": 0.8728323699421965, - "eval_runtime": 0.9503, - "eval_samples_per_second": 196.783, - "eval_steps_per_second": 3.157, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, + "eval_TIME_number": 31, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.13106106221675873, + "eval_overall_accuracy": 0.9593757740896706, + "eval_overall_f1": 0.8084544253632759, + "eval_overall_precision": 0.7445255474452555, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.3742, + "eval_samples_per_second": 499.683, + "eval_steps_per_second": 8.016, "step": 4346 }, { "epoch": 42.0, - "grad_norm": 1.8241137266159058, + "grad_norm": 1.6630405187606812, "learning_rate": 2.9e-05, - "loss": 0.1092, + "loss": 0.104, "step": 4452 }, { "epoch": 42.0, - "eval_LOCATION_f1": 0.7814569536423841, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7195121951219512, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.656934306569343, + "eval_ORGANIZATION_f1": 0.684931506849315, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5921052631578947, - "eval_ORGANIZATION_recall": 0.7377049180327869, - "eval_PERSON_f1": 0.8765432098765433, + "eval_ORGANIZATION_precision": 0.5882352941176471, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8834355828220859, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8304093567251462, - "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7142857142857143, + "eval_PERSON_precision": 0.8323699421965318, + "eval_PERSON_recall": 0.9411764705882353, + "eval_QUANTITY_f1": 0.7536231884057971, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6578947368421053, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8285714285714285, + "eval_QUANTITY_precision": 0.7027027027027027, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7435897435897436, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12787660956382751, - "eval_overall_accuracy": 0.9593757740896706, - "eval_overall_f1": 0.7978723404255319, - "eval_overall_precision": 0.7389162561576355, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9615, - "eval_samples_per_second": 194.484, - "eval_steps_per_second": 3.12, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1322728544473648, + "eval_overall_accuracy": 0.9588803567005202, + "eval_overall_f1": 0.8100263852242744, + "eval_overall_precision": 0.7451456310679612, + "eval_overall_recall": 0.8872832369942196, + "eval_runtime": 0.3728, + "eval_samples_per_second": 501.676, + "eval_steps_per_second": 8.048, "step": 4452 }, { "epoch": 43.0, - "grad_norm": 1.3770560026168823, + "grad_norm": 1.8563308715820312, "learning_rate": 2.8499999999999998e-05, - "loss": 0.1076, + "loss": 0.1066, "step": 4558 }, { "epoch": 43.0, - "eval_LOCATION_f1": 0.7763157894736842, + "eval_LOCATION_f1": 0.7837837837837838, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7108433734939759, - "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6758620689655173, + "eval_LOCATION_precision": 0.7341772151898734, + "eval_LOCATION_recall": 0.8405797101449275, + "eval_ORGANIZATION_f1": 0.6944444444444445, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5833333333333334, - "eval_ORGANIZATION_recall": 0.8032786885245902, - "eval_PERSON_f1": 0.8765432098765433, + "eval_ORGANIZATION_precision": 0.6024096385542169, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8773006134969324, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8304093567251462, - "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7042253521126761, + "eval_PERSON_precision": 0.8265895953757225, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.7536231884057971, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6410256410256411, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8, + "eval_QUANTITY_precision": 0.7027027027027027, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.717948717948718, + "eval_TIME_precision": 0.7567567567567568, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.1326487511396408, - "eval_overall_accuracy": 0.9571463958384939, - "eval_overall_f1": 0.795275590551181, - "eval_overall_precision": 0.7283653846153846, - "eval_overall_recall": 0.8757225433526011, - "eval_runtime": 0.9559, - "eval_samples_per_second": 195.621, - "eval_steps_per_second": 3.138, + "eval_loss": 0.1353512555360794, + "eval_overall_accuracy": 0.9583849393113698, + "eval_overall_f1": 0.8079470198675497, + "eval_overall_precision": 0.7457212713936431, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3756, + "eval_samples_per_second": 497.827, + "eval_steps_per_second": 7.987, "step": 4558 }, { "epoch": 44.0, - "grad_norm": 1.4139273166656494, + "grad_norm": 1.7642518281936646, "learning_rate": 2.8000000000000003e-05, - "loss": 0.1059, + "loss": 0.1046, "step": 4664 }, { "epoch": 44.0, - "eval_LOCATION_f1": 0.8108108108108109, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.759493670886076, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6758620689655173, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6714285714285714, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5833333333333334, - "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_ORGANIZATION_precision": 0.5949367088607594, + "eval_ORGANIZATION_recall": 0.7704918032786885, "eval_PERSON_f1": 0.8738461538461538, "eval_PERSON_number": 153, "eval_PERSON_precision": 0.8255813953488372, "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7536231884057971, + "eval_QUANTITY_f1": 0.6956521739130435, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.7027027027027027, - "eval_QUANTITY_recall": 0.8125, - "eval_TIME_f1": 0.8, + "eval_QUANTITY_precision": 0.6486486486486487, + "eval_QUANTITY_recall": 0.75, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.717948717948718, + "eval_TIME_precision": 0.7777777777777778, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.12715336680412292, - "eval_overall_accuracy": 0.9578895219222194, - "eval_overall_f1": 0.8058124174372523, - "eval_overall_precision": 0.7420924574209246, - "eval_overall_recall": 0.8815028901734104, - "eval_runtime": 0.9629, - "eval_samples_per_second": 194.215, - "eval_steps_per_second": 3.116, + "eval_loss": 0.1240304633975029, + "eval_overall_accuracy": 0.9608620262571216, + "eval_overall_f1": 0.7999999999999999, + "eval_overall_precision": 0.7425742574257426, + "eval_overall_recall": 0.8670520231213873, + "eval_runtime": 0.3715, + "eval_samples_per_second": 503.309, + "eval_steps_per_second": 8.074, "step": 4664 }, { "epoch": 45.0, - "grad_norm": 0.8398004770278931, + "grad_norm": 1.239845871925354, "learning_rate": 2.7500000000000004e-05, - "loss": 0.1046, + "loss": 0.102, "step": 4770 }, { "epoch": 45.0, - "eval_LOCATION_f1": 0.8108108108108109, + "eval_LOCATION_f1": 0.7972972972972974, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.759493670886076, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6527777777777777, + "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6853146853146853, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5662650602409639, - "eval_ORGANIZATION_recall": 0.7704918032786885, + "eval_ORGANIZATION_precision": 0.5975609756097561, + "eval_ORGANIZATION_recall": 0.8032786885245902, "eval_PERSON_f1": 0.8765432098765433, "eval_PERSON_number": 153, "eval_PERSON_precision": 0.8304093567251462, "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7042253521126761, + "eval_QUANTITY_f1": 0.7536231884057971, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6410256410256411, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_QUANTITY_precision": 0.7027027027027027, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12998035550117493, - "eval_overall_accuracy": 0.9583849393113698, - "eval_overall_f1": 0.8015873015873016, - "eval_overall_precision": 0.7390243902439024, - "eval_overall_recall": 0.8757225433526011, - "eval_runtime": 0.9566, - "eval_samples_per_second": 195.485, - "eval_steps_per_second": 3.136, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12775897979736328, + "eval_overall_accuracy": 0.9598711914788209, + "eval_overall_f1": 0.8095872170439414, + "eval_overall_precision": 0.7506172839506173, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3743, + "eval_samples_per_second": 499.627, + "eval_steps_per_second": 8.015, "step": 4770 }, { "epoch": 46.0, - "grad_norm": 2.6791465282440186, + "grad_norm": 6.3992600440979, "learning_rate": 2.7000000000000002e-05, - "loss": 0.1037, + "loss": 0.101, "step": 4876 }, { "epoch": 46.0, - "eval_LOCATION_f1": 0.8108108108108109, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.759493670886076, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6666666666666666, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.689655172413793, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5783132530120482, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8765432098765433, + "eval_ORGANIZATION_precision": 0.5952380952380952, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8746177370030581, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8304093567251462, - "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7142857142857143, + "eval_PERSON_precision": 0.8218390804597702, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6578947368421053, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8285714285714285, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7435897435897436, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1305205523967743, - "eval_overall_accuracy": 0.9583849393113698, - "eval_overall_f1": 0.8042328042328042, - "eval_overall_precision": 0.7414634146341463, - "eval_overall_recall": 0.8786127167630058, - "eval_runtime": 0.9593, - "eval_samples_per_second": 194.944, - "eval_steps_per_second": 3.127, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1305726170539856, + "eval_overall_accuracy": 0.9596234827842457, + "eval_overall_f1": 0.8073878627968338, + "eval_overall_precision": 0.7427184466019418, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.3711, + "eval_samples_per_second": 503.952, + "eval_steps_per_second": 8.085, "step": 4876 }, { "epoch": 47.0, - "grad_norm": 1.323526382446289, + "grad_norm": 1.040471076965332, "learning_rate": 2.6500000000000004e-05, - "loss": 0.1035, + "loss": 0.1002, "step": 4982 }, { "epoch": 47.0, - "eval_LOCATION_f1": 0.7733333333333333, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7160493827160493, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6901408450704226, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6666666666666666, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.6049382716049383, - "eval_ORGANIZATION_recall": 0.8032786885245902, - "eval_PERSON_f1": 0.8765432098765433, + "eval_ORGANIZATION_precision": 0.5783132530120482, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8773006134969324, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8304093567251462, - "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.6944444444444444, + "eval_PERSON_precision": 0.8265895953757225, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.625, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8115942028985507, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7368421052631579, + "eval_TIME_precision": 0.7777777777777778, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.1306907683610916, - "eval_overall_accuracy": 0.9573941045330691, - "eval_overall_f1": 0.797886393659181, - "eval_overall_precision": 0.7347931873479319, - "eval_overall_recall": 0.8728323699421965, - "eval_runtime": 0.9656, - "eval_samples_per_second": 193.667, - "eval_steps_per_second": 3.107, + "eval_loss": 0.1305668205022812, + "eval_overall_accuracy": 0.9598711914788209, + "eval_overall_f1": 0.8042328042328042, + "eval_overall_precision": 0.7414634146341463, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3752, + "eval_samples_per_second": 498.426, + "eval_steps_per_second": 7.996, "step": 4982 }, { "epoch": 48.0, - "grad_norm": 1.5258322954177856, + "grad_norm": 2.8856351375579834, "learning_rate": 2.6000000000000002e-05, - "loss": 0.1045, + "loss": 0.0994, "step": 5088 }, { "epoch": 48.0, - "eval_LOCATION_f1": 0.7999999999999999, + "eval_LOCATION_f1": 0.8108108108108109, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7407407407407407, + "eval_LOCATION_precision": 0.759493670886076, "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6530612244897959, + "eval_ORGANIZATION_f1": 0.6482758620689655, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5581395348837209, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8738461538461538, + "eval_ORGANIZATION_precision": 0.5595238095238095, + "eval_ORGANIZATION_recall": 0.7704918032786885, + "eval_PERSON_f1": 0.8773006134969324, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8255813953488372, - "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.742857142857143, + "eval_PERSON_precision": 0.8265895953757225, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.7142857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6842105263157895, - "eval_QUANTITY_recall": 0.8125, - "eval_TIME_f1": 0.8, + "eval_QUANTITY_precision": 0.6578947368421053, + "eval_QUANTITY_recall": 0.78125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.717948717948718, + "eval_TIME_precision": 0.7777777777777778, "eval_TIME_recall": 0.9032258064516129, - "eval_loss": 0.13318440318107605, - "eval_overall_accuracy": 0.9566509784493435, - "eval_overall_f1": 0.7979002624671916, - "eval_overall_precision": 0.7307692307692307, - "eval_overall_recall": 0.8786127167630058, - "eval_runtime": 0.9888, - "eval_samples_per_second": 189.124, - "eval_steps_per_second": 3.034, + "eval_loss": 0.13150662183761597, + "eval_overall_accuracy": 0.958632648005945, + "eval_overall_f1": 0.8015873015873016, + "eval_overall_precision": 0.7390243902439024, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3732, + "eval_samples_per_second": 501.099, + "eval_steps_per_second": 8.039, "step": 5088 }, { "epoch": 49.0, - "grad_norm": 2.2215404510498047, + "grad_norm": 3.1868813037872314, "learning_rate": 2.5500000000000003e-05, - "loss": 0.103, + "loss": 0.0995, "step": 5194 }, { "epoch": 49.0, - "eval_LOCATION_f1": 0.8053691275167786, + "eval_LOCATION_f1": 0.8163265306122449, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.75, + "eval_LOCATION_precision": 0.7692307692307693, "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.6530612244897959, + "eval_ORGANIZATION_f1": 0.6853146853146853, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5581395348837209, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8703703703703703, + "eval_ORGANIZATION_precision": 0.5975609756097561, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8711656441717791, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8245614035087719, - "eval_PERSON_recall": 0.9215686274509803, - "eval_QUANTITY_f1": 0.7462686567164178, + "eval_PERSON_precision": 0.8208092485549133, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7352941176470588, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.7142857142857143, + "eval_QUANTITY_precision": 0.6944444444444444, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8285714285714285, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7435897435897436, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12893883883953094, - "eval_overall_accuracy": 0.958632648005945, - "eval_overall_f1": 0.8005284015852047, - "eval_overall_precision": 0.7372262773722628, - "eval_overall_recall": 0.8757225433526011, - "eval_runtime": 0.9653, - "eval_samples_per_second": 193.726, - "eval_steps_per_second": 3.108, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12851661443710327, + "eval_overall_accuracy": 0.9598711914788209, + "eval_overall_f1": 0.8095872170439414, + "eval_overall_precision": 0.7506172839506173, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3781, + "eval_samples_per_second": 494.575, + "eval_steps_per_second": 7.934, "step": 5194 }, { "epoch": 50.0, - "grad_norm": 1.4190187454223633, + "grad_norm": 1.8704575300216675, "learning_rate": 2.5e-05, - "loss": 0.1033, + "loss": 0.0981, "step": 5300 }, { "epoch": 50.0, - "eval_LOCATION_f1": 0.7866666666666667, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7283950617283951, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.652482269503546, + "eval_ORGANIZATION_f1": 0.6808510638297871, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.575, - "eval_ORGANIZATION_recall": 0.7540983606557377, - "eval_PERSON_f1": 0.8738461538461538, + "eval_ORGANIZATION_precision": 0.6, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8711656441717791, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8255813953488372, + "eval_PERSON_precision": 0.8208092485549133, "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7575757575757576, + "eval_QUANTITY_f1": 0.7142857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.7352941176470589, + "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12441977858543396, - "eval_overall_accuracy": 0.9591280653950953, - "eval_overall_f1": 0.8026666666666668, - "eval_overall_precision": 0.745049504950495, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9658, - "eval_samples_per_second": 193.619, - "eval_steps_per_second": 3.106, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1275041401386261, + "eval_overall_accuracy": 0.9593757740896706, + "eval_overall_f1": 0.8010610079575597, + "eval_overall_precision": 0.7401960784313726, + "eval_overall_recall": 0.8728323699421965, + "eval_runtime": 0.371, + "eval_samples_per_second": 504.005, + "eval_steps_per_second": 8.086, "step": 5300 }, { "epoch": 51.0, - "grad_norm": 2.136300563812256, + "grad_norm": 1.2837458848953247, "learning_rate": 2.45e-05, - "loss": 0.1007, + "loss": 0.0966, "step": 5406 }, { "epoch": 51.0, - "eval_LOCATION_f1": 0.7814569536423841, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7195121951219512, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6713286713286712, + "eval_ORGANIZATION_f1": 0.689655172413793, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5853658536585366, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8703703703703703, + "eval_ORGANIZATION_precision": 0.5952380952380952, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8861538461538462, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8245614035087719, - "eval_PERSON_recall": 0.9215686274509803, + "eval_PERSON_precision": 0.8372093023255814, + "eval_PERSON_recall": 0.9411764705882353, "eval_QUANTITY_f1": 0.7142857142857143, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8285714285714285, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7435897435897436, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1288469284772873, - "eval_overall_accuracy": 0.958632648005945, - "eval_overall_f1": 0.7968337730870713, - "eval_overall_precision": 0.7330097087378641, - "eval_overall_recall": 0.8728323699421965, - "eval_runtime": 0.9664, - "eval_samples_per_second": 193.502, - "eval_steps_per_second": 3.104, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12873469293117523, + "eval_overall_accuracy": 0.9598711914788209, + "eval_overall_f1": 0.8084544253632759, + "eval_overall_precision": 0.7445255474452555, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.3726, + "eval_samples_per_second": 501.893, + "eval_steps_per_second": 8.052, "step": 5406 }, { "epoch": 52.0, - "grad_norm": 1.0658990144729614, + "grad_norm": 0.7324889898300171, "learning_rate": 2.4e-05, - "loss": 0.1007, + "loss": 0.0963, "step": 5512 }, { "epoch": 52.0, - "eval_LOCATION_f1": 0.7866666666666667, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7283950617283951, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6530612244897959, + "eval_ORGANIZATION_f1": 0.6944444444444445, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5581395348837209, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8711656441717791, + "eval_ORGANIZATION_precision": 0.6024096385542169, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8746177370030581, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8208092485549133, - "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7142857142857143, + "eval_PERSON_precision": 0.8218390804597702, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6578947368421053, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8285714285714285, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7435897435897436, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.13108685612678528, - "eval_overall_accuracy": 0.9571463958384939, - "eval_overall_f1": 0.7942332896461337, - "eval_overall_precision": 0.7266187050359713, - "eval_overall_recall": 0.8757225433526011, - "eval_runtime": 0.9612, - "eval_samples_per_second": 194.545, - "eval_steps_per_second": 3.121, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12882520258426666, + "eval_overall_accuracy": 0.9601189001733961, + "eval_overall_f1": 0.8084544253632759, + "eval_overall_precision": 0.7445255474452555, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.3724, + "eval_samples_per_second": 502.128, + "eval_steps_per_second": 8.056, "step": 5512 }, { "epoch": 53.0, - "grad_norm": 1.2142080068588257, + "grad_norm": 1.6753724813461304, "learning_rate": 2.35e-05, - "loss": 0.0987, + "loss": 0.098, "step": 5618 }, { @@ -2049,115 +2049,115 @@ "eval_LOCATION_number": 69, "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6619718309859155, + "eval_ORGANIZATION_f1": 0.6853146853146853, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5802469135802469, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8703703703703703, + "eval_ORGANIZATION_precision": 0.5975609756097561, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8711656441717791, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8245614035087719, - "eval_PERSON_recall": 0.9215686274509803, - "eval_QUANTITY_f1": 0.7462686567164178, + "eval_PERSON_precision": 0.8208092485549133, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.6857142857142857, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.7142857142857143, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8656716417910448, + "eval_QUANTITY_precision": 0.631578947368421, + "eval_QUANTITY_recall": 0.75, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.8055555555555556, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12564891576766968, - "eval_overall_accuracy": 0.9596234827842457, - "eval_overall_f1": 0.8037383177570094, - "eval_overall_precision": 0.7468982630272953, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.958, - "eval_samples_per_second": 195.204, - "eval_steps_per_second": 3.132, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12833988666534424, + "eval_overall_accuracy": 0.9593757740896706, + "eval_overall_f1": 0.7999999999999999, + "eval_overall_precision": 0.7383863080684596, + "eval_overall_recall": 0.8728323699421965, + "eval_runtime": 0.3712, + "eval_samples_per_second": 503.796, + "eval_steps_per_second": 8.082, "step": 5618 }, { "epoch": 54.0, - "grad_norm": 1.4564939737319946, + "grad_norm": 1.8212159872055054, "learning_rate": 2.3000000000000003e-05, - "loss": 0.0995, + "loss": 0.095, "step": 5724 }, { "epoch": 54.0, - "eval_LOCATION_f1": 0.8163265306122449, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7692307692307693, - "eval_LOCATION_recall": 0.8695652173913043, - "eval_ORGANIZATION_f1": 0.676470588235294, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.7, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.6133333333333333, - "eval_ORGANIZATION_recall": 0.7540983606557377, - "eval_PERSON_f1": 0.875, + "eval_ORGANIZATION_precision": 0.620253164556962, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8800000000000001, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8383233532934131, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7058823529411765, + "eval_PERSON_precision": 0.8313953488372093, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.6857142857142857, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6666666666666666, + "eval_QUANTITY_precision": 0.631578947368421, "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.8484848484848486, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12378910928964615, - "eval_overall_accuracy": 0.9608620262571216, - "eval_overall_f1": 0.8092016238159675, - "eval_overall_precision": 0.7608142493638677, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 0.9595, - "eval_samples_per_second": 194.897, - "eval_steps_per_second": 3.127, + "eval_TIME_precision": 0.8, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1265900731086731, + "eval_overall_accuracy": 0.9621005697299975, + "eval_overall_f1": 0.8079999999999999, + "eval_overall_precision": 0.75, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3725, + "eval_samples_per_second": 502.02, + "eval_steps_per_second": 8.054, "step": 5724 }, { "epoch": 55.0, - "grad_norm": 1.7350727319717407, + "grad_norm": 2.103760242462158, "learning_rate": 2.25e-05, - "loss": 0.0952, + "loss": 0.0963, "step": 5830 }, { "epoch": 55.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6575342465753424, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6993006993006993, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5647058823529412, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8792569659442725, + "eval_ORGANIZATION_precision": 0.6097560975609756, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8773006134969324, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8352941176470589, - "eval_PERSON_recall": 0.9281045751633987, - "eval_QUANTITY_f1": 0.7246376811594203, + "eval_PERSON_precision": 0.8265895953757225, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6756756756756757, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8285714285714285, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7435897435897436, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12884898483753204, - "eval_overall_accuracy": 0.9588803567005202, - "eval_overall_f1": 0.797886393659181, - "eval_overall_precision": 0.7347931873479319, - "eval_overall_recall": 0.8728323699421965, - "eval_runtime": 0.9679, - "eval_samples_per_second": 193.2, - "eval_steps_per_second": 3.099, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.13048739731311798, + "eval_overall_accuracy": 0.9606143175625464, + "eval_overall_f1": 0.8095238095238094, + "eval_overall_precision": 0.7463414634146341, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.3735, + "eval_samples_per_second": 500.645, + "eval_steps_per_second": 8.032, "step": 5830 }, { "epoch": 56.0, - "grad_norm": 1.887562870979309, + "grad_norm": 1.7916220426559448, "learning_rate": 2.2000000000000003e-05, - "loss": 0.0982, + "loss": 0.0945, "step": 5936 }, { @@ -2166,37 +2166,37 @@ "eval_LOCATION_number": 69, "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6619718309859155, + "eval_ORGANIZATION_f1": 0.7194244604316549, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5802469135802469, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8730650154798762, + "eval_ORGANIZATION_precision": 0.6410256410256411, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8765432098765433, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8294117647058824, - "eval_PERSON_recall": 0.9215686274509803, + "eval_PERSON_precision": 0.8304093567251462, + "eval_PERSON_recall": 0.9281045751633987, "eval_QUANTITY_f1": 0.6857142857142857, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.631578947368421, "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.8484848484848486, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12395640462636948, - "eval_overall_accuracy": 0.9593757740896706, - "eval_overall_f1": 0.7978723404255319, - "eval_overall_precision": 0.7389162561576355, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9574, - "eval_samples_per_second": 195.33, - "eval_steps_per_second": 3.134, + "eval_TIME_precision": 0.8, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12367260456085205, + "eval_overall_accuracy": 0.9625959871191478, + "eval_overall_f1": 0.8101604278074866, + "eval_overall_precision": 0.753731343283582, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3729, + "eval_samples_per_second": 501.411, + "eval_steps_per_second": 8.044, "step": 5936 }, { "epoch": 57.0, - "grad_norm": 0.7311828136444092, + "grad_norm": 1.4273295402526855, "learning_rate": 2.15e-05, - "loss": 0.0964, + "loss": 0.0944, "step": 6042 }, { @@ -2205,310 +2205,310 @@ "eval_LOCATION_number": 69, "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6712328767123288, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5764705882352941, - "eval_ORGANIZATION_recall": 0.8032786885245902, - "eval_PERSON_f1": 0.875, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8773006134969324, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8383233532934131, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7042253521126761, + "eval_PERSON_precision": 0.8265895953757225, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6410256410256411, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12668703496456146, - "eval_overall_accuracy": 0.9591280653950953, - "eval_overall_f1": 0.7999999999999999, - "eval_overall_precision": 0.7383863080684596, - "eval_overall_recall": 0.8728323699421965, - "eval_runtime": 1.0222, - "eval_samples_per_second": 182.947, - "eval_steps_per_second": 2.935, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12745532393455505, + "eval_overall_accuracy": 0.9603666088679713, + "eval_overall_f1": 0.8105960264900662, + "eval_overall_precision": 0.7481662591687042, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.3724, + "eval_samples_per_second": 502.154, + "eval_steps_per_second": 8.056, "step": 6042 }, { "epoch": 58.0, - "grad_norm": 0.2914753556251526, + "grad_norm": 0.433877557516098, "learning_rate": 2.1e-05, - "loss": 0.0958, + "loss": 0.094, "step": 6148 }, { "epoch": 58.0, - "eval_LOCATION_f1": 0.7866666666666667, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7283950617283951, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6530612244897959, + "eval_ORGANIZATION_f1": 0.7142857142857143, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5581395348837209, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8695652173913043, + "eval_ORGANIZATION_precision": 0.6329113924050633, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8800000000000001, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8284023668639053, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.6956521739130435, + "eval_PERSON_precision": 0.8313953488372093, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.7142857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6486486486486487, - "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.8529411764705881, + "eval_QUANTITY_precision": 0.6578947368421053, + "eval_QUANTITY_recall": 0.78125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1272968202829361, - "eval_overall_accuracy": 0.9583849393113698, - "eval_overall_f1": 0.7936507936507935, - "eval_overall_precision": 0.7317073170731707, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9623, - "eval_samples_per_second": 194.317, - "eval_steps_per_second": 3.117, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12492260336875916, + "eval_overall_accuracy": 0.9621005697299975, + "eval_overall_f1": 0.8122503328894806, + "eval_overall_precision": 0.7530864197530864, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3718, + "eval_samples_per_second": 502.971, + "eval_steps_per_second": 8.069, "step": 6148 }, { "epoch": 59.0, - "grad_norm": 0.9206826090812683, + "grad_norm": 0.6837877035140991, "learning_rate": 2.05e-05, - "loss": 0.0964, + "loss": 0.0927, "step": 6254 }, { "epoch": 59.0, - "eval_LOCATION_f1": 0.7866666666666667, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7283950617283951, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6666666666666666, + "eval_ORGANIZATION_f1": 0.6993006993006993, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5875, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8695652173913043, + "eval_ORGANIZATION_precision": 0.6097560975609756, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8773006134969324, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8284023668639053, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.6956521739130435, + "eval_PERSON_precision": 0.8265895953757225, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.7536231884057971, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6486486486486487, - "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.8787878787878788, + "eval_QUANTITY_precision": 0.7027027027027027, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.8285714285714286, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12352688610553741, - "eval_overall_accuracy": 0.9591280653950953, - "eval_overall_f1": 0.7994652406417112, - "eval_overall_precision": 0.7437810945273632, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 0.9639, - "eval_samples_per_second": 194.003, - "eval_steps_per_second": 3.112, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12483343482017517, + "eval_overall_accuracy": 0.961357443646272, + "eval_overall_f1": 0.8116710875331565, + "eval_overall_precision": 0.75, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.3719, + "eval_samples_per_second": 502.848, + "eval_steps_per_second": 8.067, "step": 6254 }, { "epoch": 60.0, - "grad_norm": 0.9167794585227966, + "grad_norm": 0.7817451357841492, "learning_rate": 2e-05, - "loss": 0.096, + "loss": 0.0928, "step": 6360 }, { "epoch": 60.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6438356164383562, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6993006993006993, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5529411764705883, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8695652173913043, + "eval_ORGANIZATION_precision": 0.6097560975609756, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8773006134969324, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8284023668639053, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7042253521126761, + "eval_PERSON_precision": 0.8265895953757225, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6410256410256411, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, - "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.13006281852722168, - "eval_overall_accuracy": 0.9573941045330691, - "eval_overall_f1": 0.7899603698811096, - "eval_overall_precision": 0.7274939172749392, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 0.966, - "eval_samples_per_second": 193.591, - "eval_steps_per_second": 3.106, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.823529411764706, + "eval_TIME_number": 31, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12917159497737885, + "eval_overall_accuracy": 0.9601189001733961, + "eval_overall_f1": 0.8095238095238094, + "eval_overall_precision": 0.7463414634146341, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.372, + "eval_samples_per_second": 502.631, + "eval_steps_per_second": 8.064, "step": 6360 }, { "epoch": 61.0, - "grad_norm": 1.5752638578414917, + "grad_norm": 1.9623644351959229, "learning_rate": 1.9500000000000003e-05, - "loss": 0.095, + "loss": 0.0931, "step": 6466 }, { "epoch": 61.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6482758620689655, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6944444444444445, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5595238095238095, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8695652173913043, + "eval_ORGANIZATION_precision": 0.6024096385542169, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8711656441717791, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8284023668639053, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7246376811594203, + "eval_PERSON_precision": 0.8208092485549133, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.6956521739130435, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6756756756756757, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_QUANTITY_precision": 0.6486486486486487, + "eval_QUANTITY_recall": 0.75, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12766651809215546, - "eval_overall_accuracy": 0.9583849393113698, - "eval_overall_f1": 0.793103448275862, - "eval_overall_precision": 0.7328431372549019, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 0.9575, - "eval_samples_per_second": 195.295, - "eval_steps_per_second": 3.133, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12796209752559662, + "eval_overall_accuracy": 0.9598711914788209, + "eval_overall_f1": 0.8026490066225165, + "eval_overall_precision": 0.7408312958435208, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3707, + "eval_samples_per_second": 504.472, + "eval_steps_per_second": 8.093, "step": 6466 }, { "epoch": 62.0, - "grad_norm": 0.7826931476593018, + "grad_norm": 1.2388701438903809, "learning_rate": 1.9e-05, - "loss": 0.0941, + "loss": 0.0894, "step": 6572 }, { "epoch": 62.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6714285714285714, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6944444444444445, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5949367088607594, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8722741433021807, + "eval_ORGANIZATION_precision": 0.6024096385542169, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8773006134969324, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8333333333333334, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7246376811594203, + "eval_PERSON_precision": 0.8265895953757225, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.6956521739130435, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6756756756756757, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_QUANTITY_precision": 0.6486486486486487, + "eval_QUANTITY_recall": 0.75, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12488686293363571, - "eval_overall_accuracy": 0.9591280653950953, - "eval_overall_f1": 0.7994652406417112, - "eval_overall_precision": 0.7437810945273632, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 1.0345, - "eval_samples_per_second": 180.763, - "eval_steps_per_second": 2.9, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.127840057015419, + "eval_overall_accuracy": 0.9598711914788209, + "eval_overall_f1": 0.8052980132450333, + "eval_overall_precision": 0.7432762836185819, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3746, + "eval_samples_per_second": 499.163, + "eval_steps_per_second": 8.008, "step": 6572 }, { "epoch": 63.0, - "grad_norm": 1.1097204685211182, + "grad_norm": 2.133678913116455, "learning_rate": 1.85e-05, - "loss": 0.0941, + "loss": 0.0911, "step": 6678 }, { "epoch": 63.0, - "eval_LOCATION_f1": 0.7866666666666667, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7283950617283951, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6388888888888888, + "eval_ORGANIZATION_f1": 0.6666666666666666, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5542168674698795, - "eval_ORGANIZATION_recall": 0.7540983606557377, - "eval_PERSON_f1": 0.875, + "eval_ORGANIZATION_precision": 0.5783132530120482, + "eval_ORGANIZATION_recall": 0.7868852459016393, + "eval_PERSON_f1": 0.8827160493827161, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8383233532934131, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7164179104477612, + "eval_PERSON_precision": 0.8362573099415205, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.6666666666666667, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6857142857142857, - "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.8405797101449276, + "eval_QUANTITY_precision": 0.6216216216216216, + "eval_QUANTITY_recall": 0.71875, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12380293756723404, - "eval_overall_accuracy": 0.9593757740896706, - "eval_overall_f1": 0.7946666666666666, - "eval_overall_precision": 0.7376237623762376, - "eval_overall_recall": 0.861271676300578, - "eval_runtime": 0.9625, - "eval_samples_per_second": 194.284, - "eval_steps_per_second": 3.117, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1259499192237854, + "eval_overall_accuracy": 0.9603666088679713, + "eval_overall_f1": 0.7984084880636606, + "eval_overall_precision": 0.7377450980392157, + "eval_overall_recall": 0.869942196531792, + "eval_runtime": 0.3802, + "eval_samples_per_second": 491.903, + "eval_steps_per_second": 7.891, "step": 6678 }, { "epoch": 64.0, - "grad_norm": 2.1919615268707275, + "grad_norm": 1.4753077030181885, "learning_rate": 1.8e-05, - "loss": 0.0957, + "loss": 0.0889, "step": 6784 }, { "epoch": 64.0, - "eval_LOCATION_f1": 0.7866666666666667, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7283950617283951, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6575342465753424, + "eval_ORGANIZATION_f1": 0.6944444444444445, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5647058823529412, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8722741433021807, + "eval_ORGANIZATION_precision": 0.6024096385542169, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8800000000000001, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8333333333333334, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7246376811594203, + "eval_PERSON_precision": 0.8313953488372093, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.7142857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6756756756756757, + "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12799490988254547, - "eval_overall_accuracy": 0.9588803567005202, - "eval_overall_f1": 0.7973509933774834, - "eval_overall_precision": 0.7359413202933985, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9554, - "eval_samples_per_second": 195.723, - "eval_steps_per_second": 3.14, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12907566130161285, + "eval_overall_accuracy": 0.9596234827842457, + "eval_overall_f1": 0.8068783068783069, + "eval_overall_precision": 0.7439024390243902, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3721, + "eval_samples_per_second": 502.538, + "eval_steps_per_second": 8.062, "step": 6784 }, { "epoch": 65.0, - "grad_norm": 0.8768384456634521, + "grad_norm": 1.000364065170288, "learning_rate": 1.75e-05, - "loss": 0.0944, + "loss": 0.0884, "step": 6890 }, { @@ -2517,1405 +2517,1405 @@ "eval_LOCATION_number": 69, "eval_LOCATION_precision": 0.7283950617283951, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6620689655172414, + "eval_ORGANIZATION_f1": 0.7092198581560285, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5714285714285714, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8668730650154798, + "eval_ORGANIZATION_precision": 0.625, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8746177370030581, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8235294117647058, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7142857142857143, + "eval_PERSON_precision": 0.8218390804597702, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6578947368421053, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12723863124847412, - "eval_overall_accuracy": 0.9591280653950953, - "eval_overall_f1": 0.7952443857331573, - "eval_overall_precision": 0.732360097323601, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9567, - "eval_samples_per_second": 195.464, - "eval_steps_per_second": 3.136, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12756387889385223, + "eval_overall_accuracy": 0.9603666088679713, + "eval_overall_f1": 0.8095238095238094, + "eval_overall_precision": 0.7463414634146341, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.3751, + "eval_samples_per_second": 498.558, + "eval_steps_per_second": 7.998, "step": 6890 }, { "epoch": 66.0, - "grad_norm": 1.0975397825241089, + "grad_norm": 1.3308097124099731, "learning_rate": 1.7000000000000003e-05, - "loss": 0.091, + "loss": 0.0905, "step": 6996 }, { "epoch": 66.0, - "eval_LOCATION_f1": 0.7814569536423841, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7195121951219512, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6666666666666666, + "eval_ORGANIZATION_f1": 0.6944444444444445, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5783132530120482, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8722741433021807, + "eval_ORGANIZATION_precision": 0.6024096385542169, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8773006134969324, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8333333333333334, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.6956521739130435, + "eval_PERSON_precision": 0.8265895953757225, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6486486486486487, - "eval_QUANTITY_recall": 0.75, - "eval_TIME_f1": 0.8405797101449276, + "eval_QUANTITY_precision": 0.6756756756756757, + "eval_QUANTITY_recall": 0.78125, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1251116245985031, - "eval_overall_accuracy": 0.9591280653950953, - "eval_overall_f1": 0.7957559681697614, - "eval_overall_precision": 0.7352941176470589, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9568, - "eval_samples_per_second": 195.452, - "eval_steps_per_second": 3.136, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12825970351696014, + "eval_overall_accuracy": 0.9603666088679713, + "eval_overall_f1": 0.8068783068783069, + "eval_overall_precision": 0.7439024390243902, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3714, + "eval_samples_per_second": 503.45, + "eval_steps_per_second": 8.077, "step": 6996 }, { "epoch": 67.0, - "grad_norm": 0.8406906127929688, + "grad_norm": 1.1464766263961792, "learning_rate": 1.65e-05, - "loss": 0.0909, + "loss": 0.0884, "step": 7102 }, { "epoch": 67.0, - "eval_LOCATION_f1": 0.7814569536423841, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7195121951219512, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6714285714285714, + "eval_ORGANIZATION_f1": 0.6993006993006993, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5949367088607594, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8668730650154798, + "eval_ORGANIZATION_precision": 0.6097560975609756, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8738461538461538, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8235294117647058, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7246376811594203, + "eval_PERSON_precision": 0.8255813953488372, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6756756756756757, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12479668110609055, - "eval_overall_accuracy": 0.9588803567005202, - "eval_overall_f1": 0.7989347536617841, - "eval_overall_precision": 0.7407407407407407, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9591, - "eval_samples_per_second": 194.977, - "eval_steps_per_second": 3.128, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12948916852474213, + "eval_overall_accuracy": 0.9596234827842457, + "eval_overall_f1": 0.8079470198675497, + "eval_overall_precision": 0.7457212713936431, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3719, + "eval_samples_per_second": 502.881, + "eval_steps_per_second": 8.068, "step": 7102 }, { "epoch": 68.0, - "grad_norm": 1.9958144426345825, + "grad_norm": 0.7333919405937195, "learning_rate": 1.6000000000000003e-05, - "loss": 0.0932, + "loss": 0.0881, "step": 7208 }, { "epoch": 68.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6575342465753424, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.689655172413793, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5647058823529412, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8668730650154798, + "eval_ORGANIZATION_precision": 0.5952380952380952, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8792569659442725, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8235294117647058, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8352941176470589, + "eval_PERSON_recall": 0.9281045751633987, "eval_QUANTITY_f1": 0.7142857142857143, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8285714285714285, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7435897435897436, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1271326243877411, - "eval_overall_accuracy": 0.9578895219222194, - "eval_overall_f1": 0.7915567282321899, - "eval_overall_precision": 0.7281553398058253, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9621, - "eval_samples_per_second": 194.357, - "eval_steps_per_second": 3.118, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12930843234062195, + "eval_overall_accuracy": 0.9596234827842457, + "eval_overall_f1": 0.8052980132450333, + "eval_overall_precision": 0.7432762836185819, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3727, + "eval_samples_per_second": 501.804, + "eval_steps_per_second": 8.05, "step": 7208 }, { "epoch": 69.0, - "grad_norm": 1.375461459159851, + "grad_norm": 1.6601905822753906, "learning_rate": 1.55e-05, - "loss": 0.0922, + "loss": 0.0903, "step": 7314 }, { "epoch": 69.0, - "eval_LOCATION_f1": 0.802721088435374, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7564102564102564, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6575342465753424, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5647058823529412, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8668730650154798, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8827160493827161, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8235294117647058, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8362573099415205, + "eval_PERSON_recall": 0.934640522875817, "eval_QUANTITY_f1": 0.7142857142857143, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1263243854045868, - "eval_overall_accuracy": 0.9581372306167947, - "eval_overall_f1": 0.7973509933774834, - "eval_overall_precision": 0.7359413202933985, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9594, - "eval_samples_per_second": 194.916, - "eval_steps_per_second": 3.127, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12589633464813232, + "eval_overall_accuracy": 0.9611097349516968, + "eval_overall_f1": 0.8100929614873839, + "eval_overall_precision": 0.7493857493857494, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3735, + "eval_samples_per_second": 500.62, + "eval_steps_per_second": 8.031, "step": 7314 }, { "epoch": 70.0, - "grad_norm": 1.1578227281570435, + "grad_norm": 1.186720609664917, "learning_rate": 1.5e-05, - "loss": 0.0918, + "loss": 0.089, "step": 7420 }, { "epoch": 70.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6618705035971222, + "eval_ORGANIZATION_f1": 0.7, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5897435897435898, - "eval_ORGANIZATION_recall": 0.7540983606557377, - "eval_PERSON_f1": 0.875, + "eval_ORGANIZATION_precision": 0.620253164556962, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8765432098765433, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8383233532934131, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7462686567164178, + "eval_PERSON_precision": 0.8304093567251462, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7352941176470588, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.7142857142857143, + "eval_QUANTITY_precision": 0.6944444444444444, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.8484848484848486, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12150081992149353, - "eval_overall_accuracy": 0.9608620262571216, - "eval_overall_f1": 0.8059299191374663, - "eval_overall_precision": 0.7550505050505051, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 0.9687, - "eval_samples_per_second": 193.041, - "eval_steps_per_second": 3.097, + "eval_TIME_precision": 0.8, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12257770448923111, + "eval_overall_accuracy": 0.9628436958137231, + "eval_overall_f1": 0.8112449799196787, + "eval_overall_precision": 0.7556109725685786, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3731, + "eval_samples_per_second": 501.269, + "eval_steps_per_second": 8.042, "step": 7420 }, { "epoch": 71.0, - "grad_norm": 0.6911057829856873, + "grad_norm": 1.3465250730514526, "learning_rate": 1.45e-05, - "loss": 0.0911, + "loss": 0.0883, "step": 7526 }, { "epoch": 71.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6620689655172414, + "eval_ORGANIZATION_f1": 0.6805555555555556, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5714285714285714, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8695652173913043, + "eval_ORGANIZATION_precision": 0.5903614457831325, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8792569659442725, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8284023668639053, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8352941176470589, + "eval_PERSON_recall": 0.9281045751633987, "eval_QUANTITY_f1": 0.7142857142857143, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8285714285714285, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7435897435897436, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1274191439151764, - "eval_overall_accuracy": 0.958632648005945, - "eval_overall_f1": 0.7973509933774834, - "eval_overall_precision": 0.7359413202933985, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9607, - "eval_samples_per_second": 194.646, - "eval_steps_per_second": 3.123, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12590442597866058, + "eval_overall_accuracy": 0.9608620262571216, + "eval_overall_f1": 0.804780876494024, + "eval_overall_precision": 0.7444717444717445, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3735, + "eval_samples_per_second": 500.704, + "eval_steps_per_second": 8.033, "step": 7526 }, { "epoch": 72.0, - "grad_norm": 1.7722148895263672, + "grad_norm": 1.610948085784912, "learning_rate": 1.4000000000000001e-05, - "loss": 0.0914, + "loss": 0.0878, "step": 7632 }, { "epoch": 72.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6666666666666666, + "eval_ORGANIZATION_f1": 0.6853146853146853, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5783132530120482, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8668730650154798, + "eval_ORGANIZATION_precision": 0.5975609756097561, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8792569659442725, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8235294117647058, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8352941176470589, + "eval_PERSON_recall": 0.9281045751633987, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8285714285714285, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7435897435897436, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12830989062786102, - "eval_overall_accuracy": 0.9583849393113698, - "eval_overall_f1": 0.7984084880636606, - "eval_overall_precision": 0.7377450980392157, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9634, - "eval_samples_per_second": 194.102, - "eval_steps_per_second": 3.114, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12648038566112518, + "eval_overall_accuracy": 0.9608620262571216, + "eval_overall_f1": 0.8058510638297872, + "eval_overall_precision": 0.7463054187192119, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3794, + "eval_samples_per_second": 492.941, + "eval_steps_per_second": 7.908, "step": 7632 }, { "epoch": 73.0, - "grad_norm": 1.7078170776367188, + "grad_norm": 1.2512165307998657, "learning_rate": 1.3500000000000001e-05, - "loss": 0.0895, + "loss": 0.0866, "step": 7738 }, { "epoch": 73.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.652482269503546, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6901408450704226, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.575, - "eval_ORGANIZATION_recall": 0.7540983606557377, - "eval_PERSON_f1": 0.8777429467084639, + "eval_ORGANIZATION_precision": 0.6049382716049383, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8773006134969324, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8433734939759037, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7462686567164178, + "eval_PERSON_precision": 0.8265895953757225, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.7142857142857143, + "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12264863401651382, - "eval_overall_accuracy": 0.9598711914788209, - "eval_overall_f1": 0.8010752688172041, - "eval_overall_precision": 0.7487437185929648, - "eval_overall_recall": 0.861271676300578, - "eval_runtime": 0.9601, - "eval_samples_per_second": 194.766, - "eval_steps_per_second": 3.125, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12454748153686523, + "eval_overall_accuracy": 0.9611097349516968, + "eval_overall_f1": 0.8074369189907038, + "eval_overall_precision": 0.7469287469287469, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3746, + "eval_samples_per_second": 499.258, + "eval_steps_per_second": 8.009, "step": 7738 }, { "epoch": 74.0, - "grad_norm": 1.3398103713989258, + "grad_norm": 1.7970552444458008, "learning_rate": 1.3000000000000001e-05, - "loss": 0.0893, + "loss": 0.0859, "step": 7844 }, { "epoch": 74.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6527777777777777, + "eval_ORGANIZATION_f1": 0.6805555555555556, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5662650602409639, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8722741433021807, + "eval_ORGANIZATION_precision": 0.5903614457831325, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8693009118541033, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8333333333333334, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7246376811594203, + "eval_PERSON_precision": 0.8125, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6756756756756757, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.835820895522388, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12828786671161652, - "eval_overall_accuracy": 0.9591280653950953, - "eval_overall_f1": 0.7989347536617841, - "eval_overall_precision": 0.7407407407407407, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9583, - "eval_samples_per_second": 195.135, - "eval_steps_per_second": 3.131, + "eval_TIME_precision": 0.7777777777777778, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12762786448001862, + "eval_overall_accuracy": 0.9601189001733961, + "eval_overall_f1": 0.8036890645586299, + "eval_overall_precision": 0.738498789346247, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3742, + "eval_samples_per_second": 499.76, + "eval_steps_per_second": 8.018, "step": 7844 }, { "epoch": 75.0, - "grad_norm": 1.5878422260284424, + "grad_norm": 0.913361132144928, "learning_rate": 1.25e-05, - "loss": 0.0877, + "loss": 0.088, "step": 7950 }, { "epoch": 75.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6620689655172414, + "eval_ORGANIZATION_f1": 0.6901408450704226, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5714285714285714, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8722741433021807, + "eval_ORGANIZATION_precision": 0.6049382716049383, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8792569659442725, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8333333333333334, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7246376811594203, + "eval_PERSON_precision": 0.8352941176470589, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.742857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6756756756756757, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_QUANTITY_precision": 0.6842105263157895, + "eval_QUANTITY_recall": 0.8125, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12763577699661255, - "eval_overall_accuracy": 0.9591280653950953, - "eval_overall_f1": 0.8005319148936171, - "eval_overall_precision": 0.7413793103448276, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9602, - "eval_samples_per_second": 194.753, - "eval_steps_per_second": 3.124, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12584900856018066, + "eval_overall_accuracy": 0.961357443646272, + "eval_overall_f1": 0.8085106382978724, + "eval_overall_precision": 0.7487684729064039, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3743, + "eval_samples_per_second": 499.605, + "eval_steps_per_second": 8.015, "step": 7950 }, { "epoch": 76.0, - "grad_norm": 1.480959415435791, + "grad_norm": 1.470176339149475, "learning_rate": 1.2e-05, - "loss": 0.088, + "loss": 0.0877, "step": 8056 }, { "epoch": 76.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7814569536423841, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6713286713286712, + "eval_LOCATION_precision": 0.7195121951219512, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6944444444444445, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5853658536585366, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.875, + "eval_ORGANIZATION_precision": 0.6024096385542169, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8738461538461538, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8383233532934131, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7246376811594203, + "eval_PERSON_precision": 0.8255813953488372, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.6857142857142857, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6756756756756757, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_QUANTITY_precision": 0.631578947368421, + "eval_QUANTITY_recall": 0.75, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12569205462932587, - "eval_overall_accuracy": 0.9591280653950953, - "eval_overall_f1": 0.7999999999999999, - "eval_overall_precision": 0.7425742574257426, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9637, - "eval_samples_per_second": 194.05, - "eval_steps_per_second": 3.113, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12841203808784485, + "eval_overall_accuracy": 0.958632648005945, + "eval_overall_f1": 0.7994722955145119, + "eval_overall_precision": 0.7354368932038835, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3725, + "eval_samples_per_second": 502.075, + "eval_steps_per_second": 8.055, "step": 8056 }, { "epoch": 77.0, - "grad_norm": 0.7544510364532471, + "grad_norm": 1.8821589946746826, "learning_rate": 1.1500000000000002e-05, - "loss": 0.0899, + "loss": 0.0865, "step": 8162 }, { "epoch": 77.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.647887323943662, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6993006993006993, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5679012345679012, - "eval_ORGANIZATION_recall": 0.7540983606557377, - "eval_PERSON_f1": 0.875, + "eval_ORGANIZATION_precision": 0.6097560975609756, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8792569659442725, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8383233532934131, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7352941176470588, + "eval_PERSON_precision": 0.8352941176470589, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6944444444444444, + "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12418851256370544, - "eval_overall_accuracy": 0.958632648005945, - "eval_overall_f1": 0.7967914438502673, - "eval_overall_precision": 0.7412935323383084, - "eval_overall_recall": 0.861271676300578, - "eval_runtime": 0.9658, - "eval_samples_per_second": 193.616, - "eval_steps_per_second": 3.106, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12691272795200348, + "eval_overall_accuracy": 0.9611097349516968, + "eval_overall_f1": 0.8085106382978724, + "eval_overall_precision": 0.7487684729064039, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3754, + "eval_samples_per_second": 498.088, + "eval_steps_per_second": 7.991, "step": 8162 }, { "epoch": 78.0, - "grad_norm": 1.431129813194275, + "grad_norm": 1.1131901741027832, "learning_rate": 1.1000000000000001e-05, - "loss": 0.0896, + "loss": 0.0873, "step": 8268 }, { "epoch": 78.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6573426573426574, + "eval_ORGANIZATION_f1": 0.6901408450704226, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.573170731707317, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.875, + "eval_ORGANIZATION_precision": 0.6049382716049383, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8738461538461538, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8383233532934131, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7246376811594203, + "eval_PERSON_precision": 0.8255813953488372, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.6956521739130435, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6756756756756757, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_QUANTITY_precision": 0.6486486486486487, + "eval_QUANTITY_recall": 0.75, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12578383088111877, - "eval_overall_accuracy": 0.9593757740896706, - "eval_overall_f1": 0.8021390374331551, - "eval_overall_precision": 0.746268656716418, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9621, - "eval_samples_per_second": 194.372, - "eval_steps_per_second": 3.118, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12489111721515656, + "eval_overall_accuracy": 0.9603666088679713, + "eval_overall_f1": 0.8021248339973439, + "eval_overall_precision": 0.742014742014742, + "eval_overall_recall": 0.8728323699421965, + "eval_runtime": 0.3728, + "eval_samples_per_second": 501.583, + "eval_steps_per_second": 8.047, "step": 8268 }, { "epoch": 79.0, - "grad_norm": 1.0035464763641357, + "grad_norm": 0.7601754069328308, "learning_rate": 1.05e-05, - "loss": 0.0911, + "loss": 0.0837, "step": 8374 }, { "epoch": 79.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.647887323943662, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5679012345679012, - "eval_ORGANIZATION_recall": 0.7540983606557377, - "eval_PERSON_f1": 0.8695652173913043, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8738461538461538, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8284023668639053, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7352941176470588, + "eval_PERSON_precision": 0.8255813953488372, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.6956521739130435, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6944444444444444, - "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_QUANTITY_precision": 0.6486486486486487, + "eval_QUANTITY_recall": 0.75, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12459558248519897, - "eval_overall_accuracy": 0.958632648005945, - "eval_overall_f1": 0.7957276368491321, - "eval_overall_precision": 0.739454094292804, - "eval_overall_recall": 0.861271676300578, - "eval_runtime": 0.9543, - "eval_samples_per_second": 195.957, - "eval_steps_per_second": 3.144, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1264895647764206, + "eval_overall_accuracy": 0.9606143175625464, + "eval_overall_f1": 0.804780876494024, + "eval_overall_precision": 0.7444717444717445, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3752, + "eval_samples_per_second": 498.393, + "eval_steps_per_second": 7.996, "step": 8374 }, { "epoch": 80.0, - "grad_norm": 0.6225095987319946, + "grad_norm": 0.6998342871665955, "learning_rate": 1e-05, - "loss": 0.0913, + "loss": 0.0866, "step": 8480 }, { "epoch": 80.0, - "eval_LOCATION_f1": 0.7733333333333333, + "eval_LOCATION_f1": 0.7866666666666667, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7160493827160493, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.647887323943662, + "eval_LOCATION_precision": 0.7283950617283951, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5679012345679012, - "eval_ORGANIZATION_recall": 0.7540983606557377, - "eval_PERSON_f1": 0.875, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8827160493827161, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8383233532934131, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7352941176470588, + "eval_PERSON_precision": 0.8362573099415205, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.7142857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6944444444444444, + "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12504763901233673, - "eval_overall_accuracy": 0.958632648005945, - "eval_overall_f1": 0.7967914438502673, - "eval_overall_precision": 0.7412935323383084, - "eval_overall_recall": 0.861271676300578, - "eval_runtime": 0.9647, - "eval_samples_per_second": 193.848, - "eval_steps_per_second": 3.11, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.13076122105121613, + "eval_overall_accuracy": 0.9603666088679713, + "eval_overall_f1": 0.8090185676392574, + "eval_overall_precision": 0.7475490196078431, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3724, + "eval_samples_per_second": 502.187, + "eval_steps_per_second": 8.056, "step": 8480 }, { "epoch": 81.0, - "grad_norm": 0.8420107364654541, + "grad_norm": 0.47672829031944275, "learning_rate": 9.5e-06, - "loss": 0.0881, + "loss": 0.084, "step": 8586 }, { "epoch": 81.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6713286713286712, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5853658536585366, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.875, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8765432098765433, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8383233532934131, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7246376811594203, + "eval_PERSON_precision": 0.8304093567251462, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7142857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6756756756756757, + "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12701000273227692, - "eval_overall_accuracy": 0.9593757740896706, - "eval_overall_f1": 0.7999999999999999, - "eval_overall_precision": 0.7425742574257426, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9588, - "eval_samples_per_second": 195.027, - "eval_steps_per_second": 3.129, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12743818759918213, + "eval_overall_accuracy": 0.9606143175625464, + "eval_overall_f1": 0.8074369189907038, + "eval_overall_precision": 0.7469287469287469, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3752, + "eval_samples_per_second": 498.348, + "eval_steps_per_second": 7.995, "step": 8586 }, { "epoch": 82.0, - "grad_norm": 0.9231528043746948, + "grad_norm": 1.3104524612426758, "learning_rate": 9e-06, - "loss": 0.0886, + "loss": 0.0858, "step": 8692 }, { "epoch": 82.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6619718309859155, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5802469135802469, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.875, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8765432098765433, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8383233532934131, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7352941176470588, + "eval_PERSON_precision": 0.8304093567251462, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6944444444444444, + "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12471184879541397, - "eval_overall_accuracy": 0.9593757740896706, - "eval_overall_f1": 0.8005354752342703, - "eval_overall_precision": 0.7456359102244389, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 0.9606, - "eval_samples_per_second": 194.672, - "eval_steps_per_second": 3.123, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12744171917438507, + "eval_overall_accuracy": 0.961357443646272, + "eval_overall_f1": 0.8085106382978724, + "eval_overall_precision": 0.7487684729064039, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3751, + "eval_samples_per_second": 498.519, + "eval_steps_per_second": 7.998, "step": 8692 }, { "epoch": 83.0, - "grad_norm": 0.5154955387115479, + "grad_norm": 0.5750565528869629, "learning_rate": 8.500000000000002e-06, - "loss": 0.0864, + "loss": 0.0837, "step": 8798 }, { "epoch": 83.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6666666666666666, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5875, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.875, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8765432098765433, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8383233532934131, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7352941176470588, + "eval_PERSON_precision": 0.8304093567251462, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6944444444444444, + "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12483326345682144, - "eval_overall_accuracy": 0.9593757740896706, - "eval_overall_f1": 0.8005354752342703, - "eval_overall_precision": 0.7456359102244389, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 0.9607, - "eval_samples_per_second": 194.657, - "eval_steps_per_second": 3.123, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12583325803279877, + "eval_overall_accuracy": 0.961357443646272, + "eval_overall_f1": 0.8085106382978724, + "eval_overall_precision": 0.7487684729064039, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3729, + "eval_samples_per_second": 501.535, + "eval_steps_per_second": 8.046, "step": 8798 }, { "epoch": 84.0, - "grad_norm": 1.965425729751587, + "grad_norm": 1.7209677696228027, "learning_rate": 8.000000000000001e-06, - "loss": 0.0882, + "loss": 0.0866, "step": 8904 }, { "epoch": 84.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.676056338028169, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6901408450704226, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5925925925925926, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.875, + "eval_ORGANIZATION_precision": 0.6049382716049383, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8765432098765433, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8383233532934131, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7352941176470588, + "eval_PERSON_precision": 0.8304093567251462, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6944444444444444, + "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12536722421646118, - "eval_overall_accuracy": 0.9596234827842457, - "eval_overall_f1": 0.8032128514056226, - "eval_overall_precision": 0.7481296758104738, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9586, - "eval_samples_per_second": 195.082, - "eval_steps_per_second": 3.13, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1269911527633667, + "eval_overall_accuracy": 0.9608620262571216, + "eval_overall_f1": 0.8058510638297872, + "eval_overall_precision": 0.7463054187192119, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3722, + "eval_samples_per_second": 502.452, + "eval_steps_per_second": 8.061, "step": 8904 }, { "epoch": 85.0, - "grad_norm": 1.8783724308013916, + "grad_norm": 1.8824012279510498, "learning_rate": 7.5e-06, - "loss": 0.0874, + "loss": 0.0837, "step": 9010 }, { "epoch": 85.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6666666666666666, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.6901408450704226, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5875, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.875, + "eval_ORGANIZATION_precision": 0.6049382716049383, + "eval_ORGANIZATION_recall": 0.8032786885245902, + "eval_PERSON_f1": 0.8765432098765433, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8383233532934131, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7352941176470588, + "eval_PERSON_precision": 0.8304093567251462, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6944444444444444, + "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1244855672121048, - "eval_overall_accuracy": 0.9596234827842457, - "eval_overall_f1": 0.8016085790884719, - "eval_overall_precision": 0.7475, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 0.9544, - "eval_samples_per_second": 195.941, - "eval_steps_per_second": 3.143, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12705950438976288, + "eval_overall_accuracy": 0.9606143175625464, + "eval_overall_f1": 0.8058510638297872, + "eval_overall_precision": 0.7463054187192119, + "eval_overall_recall": 0.8757225433526011, + "eval_runtime": 0.3715, + "eval_samples_per_second": 503.358, + "eval_steps_per_second": 8.075, "step": 9010 }, { "epoch": 86.0, - "grad_norm": 1.9328482151031494, + "grad_norm": 2.3330228328704834, "learning_rate": 7.000000000000001e-06, - "loss": 0.0876, + "loss": 0.085, "step": 9116 }, { "epoch": 86.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7866666666666667, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7283950617283951, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.676056338028169, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5925925925925926, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8757763975155279, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8827160493827161, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.834319526627219, - "eval_PERSON_recall": 0.9215686274509803, + "eval_PERSON_precision": 0.8362573099415205, + "eval_PERSON_recall": 0.934640522875817, "eval_QUANTITY_f1": 0.7142857142857143, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12838461995124817, - "eval_overall_accuracy": 0.9588803567005202, - "eval_overall_f1": 0.8042609853528628, - "eval_overall_precision": 0.745679012345679, - "eval_overall_recall": 0.8728323699421965, - "eval_runtime": 1.047, - "eval_samples_per_second": 178.61, - "eval_steps_per_second": 2.865, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12910060584545135, + "eval_overall_accuracy": 0.9606143175625464, + "eval_overall_f1": 0.8090185676392574, + "eval_overall_precision": 0.7475490196078431, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3729, + "eval_samples_per_second": 501.432, + "eval_steps_per_second": 8.044, "step": 9116 }, { "epoch": 87.0, - "grad_norm": 2.376732587814331, + "grad_norm": 2.0060434341430664, "learning_rate": 6.5000000000000004e-06, - "loss": 0.0875, + "loss": 0.0818, "step": 9222 }, { "epoch": 87.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.676056338028169, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5925925925925926, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8785046728971961, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8711656441717791, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8392857142857143, - "eval_PERSON_recall": 0.9215686274509803, + "eval_PERSON_precision": 0.8208092485549133, + "eval_PERSON_recall": 0.9281045751633987, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1258898824453354, - "eval_overall_accuracy": 0.9591280653950953, - "eval_overall_f1": 0.8026666666666668, - "eval_overall_precision": 0.745049504950495, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9634, - "eval_samples_per_second": 194.111, - "eval_steps_per_second": 3.114, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1290162354707718, + "eval_overall_accuracy": 0.9606143175625464, + "eval_overall_f1": 0.8063660477453581, + "eval_overall_precision": 0.7450980392156863, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3722, + "eval_samples_per_second": 502.389, + "eval_steps_per_second": 8.06, "step": 9222 }, { "epoch": 88.0, - "grad_norm": 0.9929248094558716, + "grad_norm": 1.1834133863449097, "learning_rate": 6e-06, - "loss": 0.0859, + "loss": 0.0814, "step": 9328 }, { "epoch": 88.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6762589928057554, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.6025641025641025, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8695652173913043, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8765432098765433, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8284023668639053, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7352941176470588, + "eval_PERSON_precision": 0.8304093567251462, + "eval_PERSON_recall": 0.9281045751633987, + "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6944444444444444, + "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12374947220087051, - "eval_overall_accuracy": 0.9593757740896706, - "eval_overall_f1": 0.8016085790884719, - "eval_overall_precision": 0.7475, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 0.954, - "eval_samples_per_second": 196.01, - "eval_steps_per_second": 3.145, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12794728577136993, + "eval_overall_accuracy": 0.9611097349516968, + "eval_overall_f1": 0.8085106382978724, + "eval_overall_precision": 0.7487684729064039, + "eval_overall_recall": 0.8786127167630058, + "eval_runtime": 0.3732, + "eval_samples_per_second": 501.069, + "eval_steps_per_second": 8.039, "step": 9328 }, { "epoch": 89.0, - "grad_norm": 0.7865117788314819, + "grad_norm": 1.2125024795532227, "learning_rate": 5.500000000000001e-06, - "loss": 0.086, + "loss": 0.0845, "step": 9434 }, { "epoch": 89.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6901408450704226, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.7132867132867133, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.6049382716049383, - "eval_ORGANIZATION_recall": 0.8032786885245902, - "eval_PERSON_f1": 0.8668730650154798, + "eval_ORGANIZATION_precision": 0.6219512195121951, + "eval_ORGANIZATION_recall": 0.8360655737704918, + "eval_PERSON_f1": 0.8827160493827161, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8235294117647058, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8362573099415205, + "eval_PERSON_recall": 0.934640522875817, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8405797101449276, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7631578947368421, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12529782950878143, - "eval_overall_accuracy": 0.9588803567005202, - "eval_overall_f1": 0.8005319148936171, - "eval_overall_precision": 0.7413793103448276, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9714, - "eval_samples_per_second": 192.499, - "eval_steps_per_second": 3.088, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1287134289741516, + "eval_overall_accuracy": 0.9616051523408472, + "eval_overall_f1": 0.8127490039840638, + "eval_overall_precision": 0.7518427518427518, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.3745, + "eval_samples_per_second": 499.313, + "eval_steps_per_second": 8.01, "step": 9434 }, { "epoch": 90.0, - "grad_norm": 0.8573620319366455, + "grad_norm": 3.6503348350524902, "learning_rate": 5e-06, - "loss": 0.0866, + "loss": 0.0838, "step": 9540 }, { "epoch": 90.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6901408450704226, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.7132867132867133, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.6049382716049383, - "eval_ORGANIZATION_recall": 0.8032786885245902, - "eval_PERSON_f1": 0.8695652173913043, + "eval_ORGANIZATION_precision": 0.6219512195121951, + "eval_ORGANIZATION_recall": 0.8360655737704918, + "eval_PERSON_f1": 0.8854489164086689, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8284023668639053, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8411764705882353, + "eval_PERSON_recall": 0.934640522875817, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12555372714996338, - "eval_overall_accuracy": 0.9591280653950953, - "eval_overall_f1": 0.8026666666666668, - "eval_overall_precision": 0.745049504950495, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9659, - "eval_samples_per_second": 193.605, - "eval_steps_per_second": 3.106, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12891411781311035, + "eval_overall_accuracy": 0.9618528610354223, + "eval_overall_f1": 0.8138297872340425, + "eval_overall_precision": 0.7536945812807881, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.3711, + "eval_samples_per_second": 503.843, + "eval_steps_per_second": 8.083, "step": 9540 }, { "epoch": 91.0, - "grad_norm": 0.8298404812812805, + "grad_norm": 0.752189040184021, "learning_rate": 4.5e-06, - "loss": 0.0867, + "loss": 0.0832, "step": 9646 }, { "epoch": 91.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6808510638297871, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.7132867132867133, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.6, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8695652173913043, + "eval_ORGANIZATION_precision": 0.6219512195121951, + "eval_ORGANIZATION_recall": 0.8360655737704918, + "eval_PERSON_f1": 0.8854489164086689, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8284023668639053, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8411764705882353, + "eval_PERSON_recall": 0.934640522875817, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12434305250644684, - "eval_overall_accuracy": 0.9588803567005202, - "eval_overall_f1": 0.801068090787717, - "eval_overall_precision": 0.7444168734491315, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9857, - "eval_samples_per_second": 189.722, - "eval_steps_per_second": 3.044, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12811307609081268, + "eval_overall_accuracy": 0.9616051523408472, + "eval_overall_f1": 0.8138297872340425, + "eval_overall_precision": 0.7536945812807881, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.372, + "eval_samples_per_second": 502.739, + "eval_steps_per_second": 8.065, "step": 9646 }, { "epoch": 92.0, - "grad_norm": 4.414638996124268, + "grad_norm": 1.4708694219589233, "learning_rate": 4.000000000000001e-06, - "loss": 0.0892, + "loss": 0.0818, "step": 9752 }, { "epoch": 92.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.6714285714285714, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5949367088607594, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8722741433021807, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8854489164086689, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8333333333333334, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8411764705882353, + "eval_PERSON_recall": 0.934640522875817, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12473282963037491, - "eval_overall_accuracy": 0.9588803567005202, - "eval_overall_f1": 0.8005354752342703, - "eval_overall_precision": 0.7456359102244389, - "eval_overall_recall": 0.8641618497109826, - "eval_runtime": 0.9575, - "eval_samples_per_second": 195.298, - "eval_steps_per_second": 3.133, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12748010456562042, + "eval_overall_accuracy": 0.9616051523408472, + "eval_overall_f1": 0.8122503328894806, + "eval_overall_precision": 0.7530864197530864, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3742, + "eval_samples_per_second": 499.78, + "eval_steps_per_second": 8.018, "step": 9752 }, { "epoch": 93.0, - "grad_norm": 0.7473860383033752, + "grad_norm": 0.8394293189048767, "learning_rate": 3.5000000000000004e-06, - "loss": 0.0859, + "loss": 0.0843, "step": 9858 }, { "epoch": 93.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6713286713286712, + "eval_ORGANIZATION_f1": 0.7083333333333334, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5853658536585366, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8722741433021807, + "eval_ORGANIZATION_precision": 0.6144578313253012, + "eval_ORGANIZATION_recall": 0.8360655737704918, + "eval_PERSON_f1": 0.8827160493827161, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8333333333333334, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7246376811594203, + "eval_PERSON_precision": 0.8362573099415205, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.7142857142857143, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6756756756756757, + "eval_QUANTITY_precision": 0.6578947368421053, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12541848421096802, - "eval_overall_accuracy": 0.9596234827842457, - "eval_overall_f1": 0.8037383177570094, - "eval_overall_precision": 0.7468982630272953, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9559, - "eval_samples_per_second": 195.623, - "eval_steps_per_second": 3.138, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12908503413200378, + "eval_overall_accuracy": 0.9608620262571216, + "eval_overall_f1": 0.8105960264900662, + "eval_overall_precision": 0.7481662591687042, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.3732, + "eval_samples_per_second": 501.076, + "eval_steps_per_second": 8.039, "step": 9858 }, { "epoch": 94.0, - "grad_norm": 3.37979793548584, + "grad_norm": 0.8784964084625244, "learning_rate": 3e-06, - "loss": 0.0848, + "loss": 0.0821, "step": 9964 }, { "epoch": 94.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.676056338028169, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5925925925925926, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8695652173913043, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8827160493827161, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8284023668639053, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8362573099415205, + "eval_PERSON_recall": 0.934640522875817, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12489689886569977, - "eval_overall_accuracy": 0.9593757740896706, - "eval_overall_f1": 0.8037383177570094, - "eval_overall_precision": 0.7468982630272953, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9637, - "eval_samples_per_second": 194.048, - "eval_steps_per_second": 3.113, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12788699567317963, + "eval_overall_accuracy": 0.961357443646272, + "eval_overall_f1": 0.8111702127659576, + "eval_overall_precision": 0.7512315270935961, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3742, + "eval_samples_per_second": 499.732, + "eval_steps_per_second": 8.017, "step": 9964 }, { "epoch": 95.0, - "grad_norm": 0.635157585144043, + "grad_norm": 1.3944754600524902, "learning_rate": 2.5e-06, - "loss": 0.0864, + "loss": 0.0807, "step": 10070 }, { "epoch": 95.0, - "eval_LOCATION_f1": 0.778523489932886, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.725, - "eval_LOCATION_recall": 0.8405797101449275, - "eval_ORGANIZATION_f1": 0.676056338028169, + "eval_LOCATION_precision": 0.7375, + "eval_LOCATION_recall": 0.855072463768116, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5925925925925926, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8722741433021807, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8827160493827161, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8333333333333334, - "eval_PERSON_recall": 0.9150326797385621, - "eval_QUANTITY_f1": 0.7352941176470588, + "eval_PERSON_precision": 0.8362573099415205, + "eval_PERSON_recall": 0.934640522875817, + "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, - "eval_QUANTITY_precision": 0.6944444444444444, + "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12473416328430176, - "eval_overall_accuracy": 0.9596234827842457, - "eval_overall_f1": 0.8021390374331551, - "eval_overall_precision": 0.746268656716418, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9585, - "eval_samples_per_second": 195.103, - "eval_steps_per_second": 3.13, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12791010737419128, + "eval_overall_accuracy": 0.9616051523408472, + "eval_overall_f1": 0.8111702127659576, + "eval_overall_precision": 0.7512315270935961, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3722, + "eval_samples_per_second": 502.394, + "eval_steps_per_second": 8.06, "step": 10070 }, { "epoch": 96.0, - "grad_norm": 1.3519235849380493, + "grad_norm": 2.0656416416168213, "learning_rate": 2.0000000000000003e-06, - "loss": 0.0879, + "loss": 0.0833, "step": 10176 }, { "epoch": 96.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.676056338028169, + "eval_ORGANIZATION_f1": 0.7132867132867133, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5925925925925926, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8695652173913043, + "eval_ORGANIZATION_precision": 0.6219512195121951, + "eval_ORGANIZATION_recall": 0.8360655737704918, + "eval_PERSON_f1": 0.8827160493827161, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8284023668639053, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8362573099415205, + "eval_PERSON_recall": 0.934640522875817, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12507809698581696, - "eval_overall_accuracy": 0.9593757740896706, - "eval_overall_f1": 0.8037383177570094, - "eval_overall_precision": 0.7468982630272953, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9561, - "eval_samples_per_second": 195.578, - "eval_steps_per_second": 3.138, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12910382449626923, + "eval_overall_accuracy": 0.961357443646272, + "eval_overall_f1": 0.8127490039840638, + "eval_overall_precision": 0.7518427518427518, + "eval_overall_recall": 0.884393063583815, + "eval_runtime": 0.3791, + "eval_samples_per_second": 493.238, + "eval_steps_per_second": 7.913, "step": 10176 }, { "epoch": 97.0, - "grad_norm": 0.8386922478675842, + "grad_norm": 6.939116477966309, "learning_rate": 1.5e-06, - "loss": 0.0858, + "loss": 0.0835, "step": 10282 }, { "epoch": 97.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.676056338028169, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5925925925925926, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8722741433021807, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8827160493827161, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8333333333333334, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8362573099415205, + "eval_PERSON_recall": 0.934640522875817, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12496434897184372, - "eval_overall_accuracy": 0.9596234827842457, - "eval_overall_f1": 0.8048128342245989, - "eval_overall_precision": 0.7487562189054726, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.96, - "eval_samples_per_second": 194.782, - "eval_steps_per_second": 3.125, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.1281178891658783, + "eval_overall_accuracy": 0.961357443646272, + "eval_overall_f1": 0.8111702127659576, + "eval_overall_precision": 0.7512315270935961, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.376, + "eval_samples_per_second": 497.302, + "eval_steps_per_second": 7.978, "step": 10282 }, { "epoch": 98.0, - "grad_norm": 0.9604553580284119, + "grad_norm": 1.4033215045928955, "learning_rate": 1.0000000000000002e-06, - "loss": 0.0857, + "loss": 0.081, "step": 10388 }, { "epoch": 98.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.676056338028169, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5925925925925926, - "eval_ORGANIZATION_recall": 0.7868852459016393, - "eval_PERSON_f1": 0.8695652173913043, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8827160493827161, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8284023668639053, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8362573099415205, + "eval_PERSON_recall": 0.934640522875817, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.12475451081991196, - "eval_overall_accuracy": 0.9593757740896706, - "eval_overall_f1": 0.8037383177570094, - "eval_overall_precision": 0.7468982630272953, - "eval_overall_recall": 0.869942196531792, - "eval_runtime": 0.9513, - "eval_samples_per_second": 196.564, - "eval_steps_per_second": 3.153, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12827619910240173, + "eval_overall_accuracy": 0.961357443646272, + "eval_overall_f1": 0.8111702127659576, + "eval_overall_precision": 0.7512315270935961, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3743, + "eval_samples_per_second": 499.639, + "eval_steps_per_second": 8.016, "step": 10388 }, { "epoch": 99.0, - "grad_norm": 1.010923147201538, + "grad_norm": 1.0604668855667114, "learning_rate": 5.000000000000001e-07, - "loss": 0.0862, + "loss": 0.0815, "step": 10494 }, { "epoch": 99.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6666666666666666, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5875, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8722741433021807, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8854489164086689, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8333333333333334, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8411764705882353, + "eval_PERSON_recall": 0.934640522875817, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1243559792637825, - "eval_overall_accuracy": 0.9593757740896706, - "eval_overall_f1": 0.8032128514056226, - "eval_overall_precision": 0.7481296758104738, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.9561, - "eval_samples_per_second": 195.582, - "eval_steps_per_second": 3.138, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12803325057029724, + "eval_overall_accuracy": 0.9616051523408472, + "eval_overall_f1": 0.8122503328894806, + "eval_overall_precision": 0.7530864197530864, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3701, + "eval_samples_per_second": 505.209, + "eval_steps_per_second": 8.105, "step": 10494 }, { "epoch": 100.0, - "grad_norm": 1.853456735610962, + "grad_norm": 1.724523901939392, "learning_rate": 0.0, - "loss": 0.0838, + "loss": 0.0848, "step": 10600 }, { "epoch": 100.0, - "eval_LOCATION_f1": 0.7972972972972974, + "eval_LOCATION_f1": 0.7919463087248322, "eval_LOCATION_number": 69, - "eval_LOCATION_precision": 0.7468354430379747, + "eval_LOCATION_precision": 0.7375, "eval_LOCATION_recall": 0.855072463768116, - "eval_ORGANIZATION_f1": 0.6666666666666666, + "eval_ORGANIZATION_f1": 0.704225352112676, "eval_ORGANIZATION_number": 61, - "eval_ORGANIZATION_precision": 0.5875, - "eval_ORGANIZATION_recall": 0.7704918032786885, - "eval_PERSON_f1": 0.8722741433021807, + "eval_ORGANIZATION_precision": 0.6172839506172839, + "eval_ORGANIZATION_recall": 0.819672131147541, + "eval_PERSON_f1": 0.8827160493827161, "eval_PERSON_number": 153, - "eval_PERSON_precision": 0.8333333333333334, - "eval_PERSON_recall": 0.9150326797385621, + "eval_PERSON_precision": 0.8362573099415205, + "eval_PERSON_recall": 0.934640522875817, "eval_QUANTITY_f1": 0.7246376811594203, "eval_QUANTITY_number": 32, "eval_QUANTITY_precision": 0.6756756756756757, "eval_QUANTITY_recall": 0.78125, - "eval_TIME_f1": 0.8529411764705881, + "eval_TIME_f1": 0.823529411764706, "eval_TIME_number": 31, - "eval_TIME_precision": 0.7837837837837838, - "eval_TIME_recall": 0.9354838709677419, - "eval_loss": 0.1244937852025032, - "eval_overall_accuracy": 0.9593757740896706, - "eval_overall_f1": 0.8032128514056226, - "eval_overall_precision": 0.7481296758104738, - "eval_overall_recall": 0.8670520231213873, - "eval_runtime": 0.955, - "eval_samples_per_second": 195.803, - "eval_steps_per_second": 3.141, + "eval_TIME_precision": 0.7567567567567568, + "eval_TIME_recall": 0.9032258064516129, + "eval_loss": 0.12820416688919067, + "eval_overall_accuracy": 0.961357443646272, + "eval_overall_f1": 0.8111702127659576, + "eval_overall_precision": 0.7512315270935961, + "eval_overall_recall": 0.8815028901734104, + "eval_runtime": 0.3721, + "eval_samples_per_second": 502.522, + "eval_steps_per_second": 8.062, "step": 10600 }, { "epoch": 100.0, "step": 10600, "total_flos": 4576320229644174.0, - "train_loss": 0.14507540163004173, - "train_runtime": 1241.1304, - "train_samples_per_second": 135.924, - "train_steps_per_second": 8.541 + "train_loss": 0.14307242384496724, + "train_runtime": 560.7825, + "train_samples_per_second": 300.83, + "train_steps_per_second": 18.902 } ], "logging_steps": 500,