| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 31.0, | |
| "eval_steps": 500, | |
| "global_step": 2511, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_B": { | |
| "f1-score": 0.8087666523420713, | |
| "precision": 0.8189730200174065, | |
| "recall": 0.7988115449915111, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9374597034171502, | |
| "precision": 0.9521418826739427, | |
| "recall": 0.9232234509762421, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.8881362400419788, | |
| "precision": 0.863303347862376, | |
| "recall": 0.9144400785854617, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9154245298608586, | |
| "eval_loss": 0.2228638380765915, | |
| "eval_macro avg": { | |
| "f1-score": 0.8781208652670669, | |
| "precision": 0.8781394168512416, | |
| "recall": 0.8788250248510717, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7167, | |
| "eval_samples_per_second": 17.173, | |
| "eval_steps_per_second": 2.332, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9158543469268937, | |
| "precision": 0.9170673807540184, | |
| "recall": 0.9154245298608586, | |
| "support": 30257.0 | |
| }, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_B": { | |
| "f1-score": 0.8401515151515152, | |
| "precision": 0.7585499316005472, | |
| "recall": 0.9414261460101867, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9568235603470944, | |
| "precision": 0.9510219016256338, | |
| "recall": 0.9626964389650247, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9185648054827655, | |
| "precision": 0.9430877483443708, | |
| "recall": 0.8952848722986247, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9391876260038999, | |
| "eval_loss": 0.16692574322223663, | |
| "eval_macro avg": { | |
| "f1-score": 0.905179960327125, | |
| "precision": 0.8842198605235172, | |
| "recall": 0.9331358190912787, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7116, | |
| "eval_samples_per_second": 17.192, | |
| "eval_steps_per_second": 2.335, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9394089523635119, | |
| "precision": 0.9408589092241132, | |
| "recall": 0.9391876260038999, | |
| "support": 30257.0 | |
| }, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_B": { | |
| "f1-score": 0.8612244897959184, | |
| "precision": 0.8294025157232704, | |
| "recall": 0.8955857385398981, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.951106371634231, | |
| "precision": 0.9584654236741712, | |
| "recall": 0.9438594634636753, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9106743212999903, | |
| "precision": 0.9021592442645074, | |
| "recall": 0.9193516699410609, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9337343424662061, | |
| "eval_loss": 0.1742253601551056, | |
| "eval_macro avg": { | |
| "f1-score": 0.9076683942433799, | |
| "precision": 0.8966757278873163, | |
| "recall": 0.9195989573148781, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7094, | |
| "eval_samples_per_second": 17.2, | |
| "eval_steps_per_second": 2.336, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9340035812317422, | |
| "precision": 0.9344963252190521, | |
| "recall": 0.9337343424662061, | |
| "support": 30257.0 | |
| }, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_B": { | |
| "f1-score": 0.8717948717948719, | |
| "precision": 0.8254931714719271, | |
| "recall": 0.9235993208828522, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9561271586371415, | |
| "precision": 0.9374586871408959, | |
| "recall": 0.9755542621302714, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9117828500925355, | |
| "precision": 0.956427955133736, | |
| "recall": 0.8711198428290766, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9383944211256899, | |
| "eval_loss": 0.18792153894901276, | |
| "eval_macro avg": { | |
| "f1-score": 0.9132349601748496, | |
| "precision": 0.9064599379155197, | |
| "recall": 0.9234244752807333, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7342, | |
| "eval_samples_per_second": 17.11, | |
| "eval_steps_per_second": 2.324, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9379241479327001, | |
| "precision": 0.9394817486046585, | |
| "recall": 0.9383944211256899, | |
| "support": 30257.0 | |
| }, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_B": { | |
| "f1-score": 0.8742949234488315, | |
| "precision": 0.8320552147239264, | |
| "recall": 0.9210526315789473, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9583957528450158, | |
| "precision": 0.9521148825065274, | |
| "recall": 0.9647600402137679, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9198818996146724, | |
| "precision": 0.9375701315923697, | |
| "recall": 0.9028487229862475, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9422282447037049, | |
| "eval_loss": 0.19990424811840057, | |
| "eval_macro avg": { | |
| "f1-score": 0.9175241919695066, | |
| "precision": 0.9072467429409411, | |
| "recall": 0.9295537982596542, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7101, | |
| "eval_samples_per_second": 17.197, | |
| "eval_steps_per_second": 2.335, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9421634164298524, | |
| "precision": 0.9425469857238316, | |
| "recall": 0.9422282447037049, | |
| "support": 30257.0 | |
| }, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_B": { | |
| "f1-score": 0.8808247422680412, | |
| "precision": 0.8564554931836408, | |
| "recall": 0.9066213921901528, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.957406827435014, | |
| "precision": 0.9446361435855178, | |
| "recall": 0.9705275411397428, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9163000050574014, | |
| "precision": 0.9443344105076619, | |
| "recall": 0.8898821218074656, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9409062365733549, | |
| "eval_loss": 0.20615626871585846, | |
| "eval_macro avg": { | |
| "f1-score": 0.9181771915868189, | |
| "precision": 0.9151420157589403, | |
| "recall": 0.922343685045787, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.6949, | |
| "eval_samples_per_second": 17.253, | |
| "eval_steps_per_second": 2.343, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9405948120953971, | |
| "precision": 0.941101475610967, | |
| "recall": 0.9409062365733549, | |
| "support": 30257.0 | |
| }, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 6.17, | |
| "grad_norm": 2.4104158878326416, | |
| "learning_rate": 1.7530864197530865e-05, | |
| "loss": 0.1662, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_B": { | |
| "f1-score": 0.8836820083682009, | |
| "precision": 0.8712871287128713, | |
| "recall": 0.8964346349745331, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9614174269469787, | |
| "precision": 0.9609854091774159, | |
| "recall": 0.9618498333245146, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9271751440248166, | |
| "precision": 0.9295093296475466, | |
| "recall": 0.924852652259332, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9468552731599299, | |
| "eval_loss": 0.22925782203674316, | |
| "eval_macro avg": { | |
| "f1-score": 0.9240915264466655, | |
| "precision": 0.9205939558459445, | |
| "recall": 0.9277123735194599, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7288, | |
| "eval_samples_per_second": 17.129, | |
| "eval_steps_per_second": 2.326, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9468700903559945, | |
| "precision": 0.9469030129054358, | |
| "recall": 0.9468552731599299, | |
| "support": 30257.0 | |
| }, | |
| "step": 567 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_B": { | |
| "f1-score": 0.8803630363036303, | |
| "precision": 0.8563402889245586, | |
| "recall": 0.9057724957555179, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9580651967507121, | |
| "precision": 0.9550928116947994, | |
| "recall": 0.9610561405365363, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9208882720333101, | |
| "precision": 0.9294576746047628, | |
| "recall": 0.912475442043222, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9425587467362925, | |
| "eval_loss": 0.22973769903182983, | |
| "eval_macro avg": { | |
| "f1-score": 0.9197721683625509, | |
| "precision": 0.9136302584080402, | |
| "recall": 0.9264346927784253, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7191, | |
| "eval_samples_per_second": 17.164, | |
| "eval_steps_per_second": 2.331, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9425317916335554, | |
| "precision": 0.942623096673485, | |
| "recall": 0.9425587467362925, | |
| "support": 30257.0 | |
| }, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_B": { | |
| "f1-score": 0.8885191347753745, | |
| "precision": 0.8711256117455138, | |
| "recall": 0.9066213921901528, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9620326487625066, | |
| "precision": 0.9574445783763954, | |
| "recall": 0.9666649029049156, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9275707898658718, | |
| "precision": 0.9382914572864322, | |
| "recall": 0.9170923379174852, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9476484780381399, | |
| "eval_loss": 0.24401217699050903, | |
| "eval_macro avg": { | |
| "f1-score": 0.9260408578012509, | |
| "precision": 0.9222872158027805, | |
| "recall": 0.9301262110041845, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.701, | |
| "eval_samples_per_second": 17.23, | |
| "eval_steps_per_second": 2.34, | |
| "eval_weighted avg": { | |
| "f1-score": 0.94757580760031, | |
| "precision": 0.9476398219436029, | |
| "recall": 0.9476484780381399, | |
| "support": 30257.0 | |
| }, | |
| "step": 729 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_B": { | |
| "f1-score": 0.875959595959596, | |
| "precision": 0.8357748650732459, | |
| "recall": 0.9202037351443124, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.955652945166171, | |
| "precision": 0.9424749163879599, | |
| "recall": 0.9692047198264458, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9127632580563309, | |
| "precision": 0.9441469816272966, | |
| "recall": 0.8833988212180747, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9384274713289487, | |
| "eval_loss": 0.29471203684806824, | |
| "eval_macro avg": { | |
| "f1-score": 0.9147919330606994, | |
| "precision": 0.9074655876961675, | |
| "recall": 0.9242690920629443, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.6985, | |
| "eval_samples_per_second": 17.24, | |
| "eval_steps_per_second": 2.341, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9381199518045186, | |
| "precision": 0.9388833165494999, | |
| "recall": 0.9384274713289487, | |
| "support": 30257.0 | |
| }, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_B": { | |
| "f1-score": 0.875515251442704, | |
| "precision": 0.8509615384615384, | |
| "recall": 0.9015280135823429, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9575287538250502, | |
| "precision": 0.9547582723972855, | |
| "recall": 0.96031536060109, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9195242814667988, | |
| "precision": 0.9278, | |
| "recall": 0.9113948919449901, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9415672406385299, | |
| "eval_loss": 0.30941879749298096, | |
| "eval_macro avg": { | |
| "f1-score": 0.9175227622448511, | |
| "precision": 0.9111732702862746, | |
| "recall": 0.9244127553761411, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7017, | |
| "eval_samples_per_second": 17.228, | |
| "eval_steps_per_second": 2.34, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9415490653426031, | |
| "precision": 0.9416470001105195, | |
| "recall": 0.9415672406385299, | |
| "support": 30257.0 | |
| }, | |
| "step": 891 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_B": { | |
| "f1-score": 0.8818181818181817, | |
| "precision": 0.8590982286634461, | |
| "recall": 0.9057724957555179, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9542487087386278, | |
| "precision": 0.9457436856875585, | |
| "recall": 0.9629080903751521, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.910740239562973, | |
| "precision": 0.9297042873222142, | |
| "recall": 0.8925343811394892, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9370063125888224, | |
| "eval_loss": 0.28610122203826904, | |
| "eval_macro avg": { | |
| "f1-score": 0.9156023767065942, | |
| "precision": 0.9115154005577396, | |
| "recall": 0.9204049890900531, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.6911, | |
| "eval_samples_per_second": 17.267, | |
| "eval_steps_per_second": 2.345, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9367902899621314, | |
| "precision": 0.9369738332985704, | |
| "recall": 0.9370063125888224, | |
| "support": 30257.0 | |
| }, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 12.35, | |
| "grad_norm": 0.9797911643981934, | |
| "learning_rate": 1.506172839506173e-05, | |
| "loss": 0.0233, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_B": { | |
| "f1-score": 0.8875154511742892, | |
| "precision": 0.8622898318654924, | |
| "recall": 0.9142614601018676, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9602009576889865, | |
| "precision": 0.9497877627083549, | |
| "recall": 0.9708450182549341, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9229994967287367, | |
| "precision": 0.9463364293085655, | |
| "recall": 0.9007858546168959, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9450705621839575, | |
| "eval_loss": 0.3311145305633545, | |
| "eval_macro avg": { | |
| "f1-score": 0.9235719685306708, | |
| "precision": 0.9194713412941375, | |
| "recall": 0.9286307776578991, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7187, | |
| "eval_samples_per_second": 17.166, | |
| "eval_steps_per_second": 2.331, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9448546114137558, | |
| "precision": 0.9452199887538072, | |
| "recall": 0.9450705621839575, | |
| "support": 30257.0 | |
| }, | |
| "step": 1053 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_B": { | |
| "f1-score": 0.8758762886597938, | |
| "precision": 0.8516439454691259, | |
| "recall": 0.9015280135823429, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9588910891089109, | |
| "precision": 0.9569456155143339, | |
| "recall": 0.9608444891264089, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9223310576827941, | |
| "precision": 0.9290412597169623, | |
| "recall": 0.9157170923379175, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9433519516145025, | |
| "eval_loss": 0.32673054933547974, | |
| "eval_macro avg": { | |
| "f1-score": 0.9190328118171663, | |
| "precision": 0.9125436069001407, | |
| "recall": 0.9260298650155564, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7171, | |
| "eval_samples_per_second": 17.171, | |
| "eval_steps_per_second": 2.332, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9433584006451857, | |
| "precision": 0.943457440568685, | |
| "recall": 0.9433519516145025, | |
| "support": 30257.0 | |
| }, | |
| "step": 1134 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_B": { | |
| "f1-score": 0.8844750709363599, | |
| "precision": 0.8463925523661753, | |
| "recall": 0.9261460101867572, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9554282158321882, | |
| "precision": 0.9367500508440105, | |
| "recall": 0.974866395047357, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9105749486652978, | |
| "precision": 0.9536559139784946, | |
| "recall": 0.8712180746561886, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9380969692963612, | |
| "eval_loss": 0.4276476800441742, | |
| "eval_macro avg": { | |
| "f1-score": 0.9168260784779486, | |
| "precision": 0.9122661723962269, | |
| "recall": 0.924076826630101, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.6973, | |
| "eval_samples_per_second": 17.244, | |
| "eval_steps_per_second": 2.342, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9375748574540862, | |
| "precision": 0.9389201454833387, | |
| "recall": 0.9380969692963612, | |
| "support": 30257.0 | |
| }, | |
| "step": 1215 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_B": { | |
| "f1-score": 0.8840336134453782, | |
| "precision": 0.8752079866888519, | |
| "recall": 0.8930390492359932, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9589266737513285, | |
| "precision": 0.9629688917346994, | |
| "recall": 0.9549182496428382, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.923489801893237, | |
| "precision": 0.9174907892185379, | |
| "recall": 0.9295677799607073, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9439799054764186, | |
| "eval_loss": 0.3447544574737549, | |
| "eval_macro avg": { | |
| "f1-score": 0.9221500296966479, | |
| "precision": 0.9185558892140296, | |
| "recall": 0.9258416929465129, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7359, | |
| "eval_samples_per_second": 17.103, | |
| "eval_steps_per_second": 2.323, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9440880783665984, | |
| "precision": 0.9442509279657028, | |
| "recall": 0.9439799054764186, | |
| "support": 30257.0 | |
| }, | |
| "step": 1296 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_B": { | |
| "f1-score": 0.8751040799333889, | |
| "precision": 0.8586601307189542, | |
| "recall": 0.8921901528013583, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9534606521336053, | |
| "precision": 0.954598493688342, | |
| "recall": 0.9523255198687761, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9118326047448303, | |
| "precision": 0.9118773946360154, | |
| "recall": 0.9117878192534381, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9363453085236474, | |
| "eval_loss": 0.3998367488384247, | |
| "eval_macro avg": { | |
| "f1-score": 0.9134657789372748, | |
| "precision": 0.9083786730144373, | |
| "recall": 0.9187678306411908, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7114, | |
| "eval_samples_per_second": 17.192, | |
| "eval_steps_per_second": 2.335, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9364041837306049, | |
| "precision": 0.936489752572877, | |
| "recall": 0.9363453085236474, | |
| "support": 30257.0 | |
| }, | |
| "step": 1377 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_B": { | |
| "f1-score": 0.8807644370585792, | |
| "precision": 0.8624898291293734, | |
| "recall": 0.8998302207130731, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.958089514470979, | |
| "precision": 0.9545669415410474, | |
| "recall": 0.961638181914387, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9206207546234321, | |
| "precision": 0.9294223646010612, | |
| "recall": 0.9119842829076621, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9425256965330336, | |
| "eval_loss": 0.3564019203186035, | |
| "eval_macro avg": { | |
| "f1-score": 0.9198249020509968, | |
| "precision": 0.915493045090494, | |
| "recall": 0.9244842285117074, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.6986, | |
| "eval_samples_per_second": 17.239, | |
| "eval_steps_per_second": 2.341, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9424726021386316, | |
| "precision": 0.9425221706890129, | |
| "recall": 0.9425256965330336, | |
| "support": 30257.0 | |
| }, | |
| "step": 1458 | |
| }, | |
| { | |
| "epoch": 18.52, | |
| "grad_norm": 0.01765240728855133, | |
| "learning_rate": 1.2592592592592593e-05, | |
| "loss": 0.0067, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "eval_B": { | |
| "f1-score": 0.8756097560975611, | |
| "precision": 0.8400936037441498, | |
| "recall": 0.9142614601018676, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9563048159235334, | |
| "precision": 0.9492258770786633, | |
| "recall": 0.9634901317530028, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9161826557180053, | |
| "precision": 0.9343341503267973, | |
| "recall": 0.8987229862475442, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9397825296625574, | |
| "eval_loss": 0.41976475715637207, | |
| "eval_macro avg": { | |
| "f1-score": 0.9160324092463666, | |
| "precision": 0.9078845437165368, | |
| "recall": 0.9254915260341382, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7401, | |
| "eval_samples_per_second": 17.088, | |
| "eval_steps_per_second": 2.321, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9396639602085823, | |
| "precision": 0.9399666776761432, | |
| "recall": 0.9397825296625574, | |
| "support": 30257.0 | |
| }, | |
| "step": 1539 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_B": { | |
| "f1-score": 0.8851099128992119, | |
| "precision": 0.8653690186536902, | |
| "recall": 0.9057724957555179, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9574706357267495, | |
| "precision": 0.9427633603446507, | |
| "recall": 0.972644055241018, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9156602050137015, | |
| "precision": 0.9470921688011757, | |
| "recall": 0.8862475442043222, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9409723369798725, | |
| "eval_loss": 0.4046960175037384, | |
| "eval_macro avg": { | |
| "f1-score": 0.9194135845465543, | |
| "precision": 0.9184081825998388, | |
| "recall": 0.921554698400286, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.6993, | |
| "eval_samples_per_second": 17.237, | |
| "eval_steps_per_second": 2.341, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9405862414989786, | |
| "precision": 0.9412065878812694, | |
| "recall": 0.9409723369798725, | |
| "support": 30257.0 | |
| }, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "eval_B": { | |
| "f1-score": 0.8877928483353884, | |
| "precision": 0.8605577689243028, | |
| "recall": 0.9168081494057725, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9589888818963709, | |
| "precision": 0.9505640172584082, | |
| "recall": 0.9675644213979576, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9206317372775131, | |
| "precision": 0.9401945724526369, | |
| "recall": 0.9018664047151277, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9434841524275375, | |
| "eval_loss": 0.3954521417617798, | |
| "eval_macro avg": { | |
| "f1-score": 0.9224711558364241, | |
| "precision": 0.9171054528784492, | |
| "recall": 0.9287463251729525, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.703, | |
| "eval_samples_per_second": 17.223, | |
| "eval_steps_per_second": 2.339, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9433116944767718, | |
| "precision": 0.9435709806500092, | |
| "recall": 0.9434841524275375, | |
| "support": 30257.0 | |
| }, | |
| "step": 1701 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "eval_B": { | |
| "f1-score": 0.882280049566295, | |
| "precision": 0.8592115848753017, | |
| "recall": 0.9066213921901528, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.955072275035812, | |
| "precision": 0.9404493229380386, | |
| "recall": 0.9701571511720197, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9106508274951772, | |
| "precision": 0.9423198150872032, | |
| "recall": 0.881041257367387, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9377003668572562, | |
| "eval_loss": 0.45334869623184204, | |
| "eval_macro avg": { | |
| "f1-score": 0.9160010506990948, | |
| "precision": 0.9139935743001812, | |
| "recall": 0.9192732669098532, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7089, | |
| "eval_samples_per_second": 17.201, | |
| "eval_steps_per_second": 2.336, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9372926016522395, | |
| "precision": 0.9379158118378169, | |
| "recall": 0.9377003668572562, | |
| "support": 30257.0 | |
| }, | |
| "step": 1782 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "eval_B": { | |
| "f1-score": 0.8881632653061224, | |
| "precision": 0.8553459119496856, | |
| "recall": 0.9235993208828522, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9594668757350059, | |
| "precision": 0.947898378601673, | |
| "recall": 0.9713212339277211, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9211576342239507, | |
| "precision": 0.9480195446512112, | |
| "recall": 0.8957760314341847, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9440460058829362, | |
| "eval_loss": 0.43748563528060913, | |
| "eval_macro avg": { | |
| "f1-score": 0.9229292584216929, | |
| "precision": 0.9170879450675233, | |
| "recall": 0.9302321954149194, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.714, | |
| "eval_samples_per_second": 17.183, | |
| "eval_steps_per_second": 2.333, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9438016170620454, | |
| "precision": 0.9443357869590202, | |
| "recall": 0.9440460058829362, | |
| "support": 30257.0 | |
| }, | |
| "step": 1863 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "eval_B": { | |
| "f1-score": 0.8849630238290879, | |
| "precision": 0.857484076433121, | |
| "recall": 0.9142614601018676, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.960634026800146, | |
| "precision": 0.9468112441543759, | |
| "recall": 0.974866395047357, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9227258898691816, | |
| "precision": 0.9535736742821211, | |
| "recall": 0.8938113948919449, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9452358132002512, | |
| "eval_loss": 0.4326893985271454, | |
| "eval_macro avg": { | |
| "f1-score": 0.9227743134994718, | |
| "precision": 0.9192896649565393, | |
| "recall": 0.9276464166803898, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.6967, | |
| "eval_samples_per_second": 17.246, | |
| "eval_steps_per_second": 2.342, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9449336838891791, | |
| "precision": 0.9456086839245054, | |
| "recall": 0.9452358132002512, | |
| "support": 30257.0 | |
| }, | |
| "step": 1944 | |
| }, | |
| { | |
| "epoch": 24.69, | |
| "grad_norm": 0.07635599374771118, | |
| "learning_rate": 1.0123456790123458e-05, | |
| "loss": 0.0035, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "eval_B": { | |
| "f1-score": 0.8899958488999585, | |
| "precision": 0.8708367181153533, | |
| "recall": 0.9100169779286927, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9602969235996759, | |
| "precision": 0.9488636363636364, | |
| "recall": 0.9720091010106355, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.92240249924418, | |
| "precision": 0.946927374301676, | |
| "recall": 0.8991159135559922, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9450705621839575, | |
| "eval_loss": 0.41382941603660583, | |
| "eval_macro avg": { | |
| "f1-score": 0.9242317572479383, | |
| "precision": 0.9222092429268886, | |
| "recall": 0.9270473308317735, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.755, | |
| "eval_samples_per_second": 17.035, | |
| "eval_steps_per_second": 2.313, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9448102624655512, | |
| "precision": 0.9451743460345479, | |
| "recall": 0.9450705621839575, | |
| "support": 30257.0 | |
| }, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "eval_B": { | |
| "f1-score": 0.885489871847871, | |
| "precision": 0.863013698630137, | |
| "recall": 0.9091680814940577, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9624506968183014, | |
| "precision": 0.9566149181955987, | |
| "recall": 0.9683581141859358, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9278843757787192, | |
| "precision": 0.9417298937784522, | |
| "recall": 0.9144400785854617, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9479128796642099, | |
| "eval_loss": 0.37765103578567505, | |
| "eval_macro avg": { | |
| "f1-score": 0.9252749814816305, | |
| "precision": 0.9204528368680626, | |
| "recall": 0.9306554247551517, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7261, | |
| "eval_samples_per_second": 17.139, | |
| "eval_steps_per_second": 2.327, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9478244946172202, | |
| "precision": 0.9479626464827831, | |
| "recall": 0.9479128796642099, | |
| "support": 30257.0 | |
| }, | |
| "step": 2106 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "eval_B": { | |
| "f1-score": 0.8870500620604055, | |
| "precision": 0.8652138821630347, | |
| "recall": 0.9100169779286927, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9591804653112554, | |
| "precision": 0.949971456743993, | |
| "recall": 0.9685697655960633, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9205680164584273, | |
| "precision": 0.940917017129962, | |
| "recall": 0.9010805500982318, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9435833030373136, | |
| "eval_loss": 0.44144001603126526, | |
| "eval_macro avg": { | |
| "f1-score": 0.922266181276696, | |
| "precision": 0.9187007853456634, | |
| "recall": 0.926555764540996, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7203, | |
| "eval_samples_per_second": 17.16, | |
| "eval_steps_per_second": 2.33, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9433810025637493, | |
| "precision": 0.9436252023854246, | |
| "recall": 0.9435833030373136, | |
| "support": 30257.0 | |
| }, | |
| "step": 2187 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "eval_B": { | |
| "f1-score": 0.8864954432477217, | |
| "precision": 0.8656957928802589, | |
| "recall": 0.9083191850594228, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9565172351280657, | |
| "precision": 0.9367454600791316, | |
| "recall": 0.9771416477062278, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.912197875506748, | |
| "precision": 0.9549801224884495, | |
| "recall": 0.8730844793713163, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9394520276299699, | |
| "eval_loss": 0.5108960270881653, | |
| "eval_macro avg": { | |
| "f1-score": 0.9184035179608451, | |
| "precision": 0.9191404584826133, | |
| "recall": 0.9195151040456556, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.736, | |
| "eval_samples_per_second": 17.103, | |
| "eval_steps_per_second": 2.323, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9388797710113306, | |
| "precision": 0.9401143451426405, | |
| "recall": 0.9394520276299699, | |
| "support": 30257.0 | |
| }, | |
| "step": 2268 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "eval_B": { | |
| "f1-score": 0.8870292887029289, | |
| "precision": 0.8745874587458746, | |
| "recall": 0.8998302207130731, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9617674295541738, | |
| "precision": 0.9618437764606266, | |
| "recall": 0.9616910947669188, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9281322249003886, | |
| "precision": 0.9295497093309686, | |
| "recall": 0.9267190569744598, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9475162772251049, | |
| "eval_loss": 0.41041621565818787, | |
| "eval_macro avg": { | |
| "f1-score": 0.9256429810524972, | |
| "precision": 0.9219936481791565, | |
| "recall": 0.9294134574848173, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7029, | |
| "eval_samples_per_second": 17.224, | |
| "eval_steps_per_second": 2.339, | |
| "eval_weighted avg": { | |
| "f1-score": 0.94754103852736, | |
| "precision": 0.9475812406623685, | |
| "recall": 0.9475162772251049, | |
| "support": 30257.0 | |
| }, | |
| "step": 2349 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "eval_B": { | |
| "f1-score": 0.885939444214019, | |
| "precision": 0.8661800486618005, | |
| "recall": 0.9066213921901528, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.959851047649018, | |
| "precision": 0.9514921493189145, | |
| "recall": 0.9683581141859358, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9222834251377066, | |
| "precision": 0.9406537282941777, | |
| "recall": 0.9046168958742633, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9445087087285586, | |
| "eval_loss": 0.4500040113925934, | |
| "eval_macro avg": { | |
| "f1-score": 0.9226913056669145, | |
| "precision": 0.9194419754249642, | |
| "recall": 0.9265321340834506, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7043, | |
| "eval_samples_per_second": 17.218, | |
| "eval_steps_per_second": 2.338, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9443337701260784, | |
| "precision": 0.9445240830662821, | |
| "recall": 0.9445087087285586, | |
| "support": 30257.0 | |
| }, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 30.86, | |
| "grad_norm": 0.011806252412497997, | |
| "learning_rate": 7.654320987654322e-06, | |
| "loss": 0.002, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "eval_B": { | |
| "f1-score": 0.8837792642140467, | |
| "precision": 0.8706754530477759, | |
| "recall": 0.8972835314091681, | |
| "support": 1178.0 | |
| }, | |
| "eval_I": { | |
| "f1-score": 0.9593350046180235, | |
| "precision": 0.9568856601389766, | |
| "recall": 0.9617969204719826, | |
| "support": 18899.0 | |
| }, | |
| "eval_O": { | |
| "f1-score": 0.9225292925297869, | |
| "precision": 0.9286354135562854, | |
| "recall": 0.9165029469548134, | |
| "support": 10180.0 | |
| }, | |
| "eval_accuracy": 0.9440460058829362, | |
| "eval_loss": 0.4298489987850189, | |
| "eval_macro avg": { | |
| "f1-score": 0.9218811871206191, | |
| "precision": 0.9187321755810127, | |
| "recall": 0.9251944662786548, | |
| "support": 30257.0 | |
| }, | |
| "eval_runtime": 4.7086, | |
| "eval_samples_per_second": 17.203, | |
| "eval_steps_per_second": 2.336, | |
| "eval_weighted avg": { | |
| "f1-score": 0.9440100612576726, | |
| "precision": 0.9440244004580688, | |
| "recall": 0.9440460058829362, | |
| "support": 30257.0 | |
| }, | |
| "step": 2511 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 4050, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 50, | |
| "save_steps": 500, | |
| "total_flos": 4443543420409800.0, | |
| "train_batch_size": 4, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |