| { | |
| "best_metric": 0.6460817297487148, | |
| "best_model_checkpoint": "InLegalBERT-lora/checkpoint-2400", | |
| "epoch": 14.930015552099533, | |
| "eval_steps": 500, | |
| "global_step": 2400, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.6553059643687065, | |
| "eval_f1_macro": 0.34196434919671853, | |
| "eval_f1_micro": 0.6553059643687065, | |
| "eval_f1_weighted": 0.6146575775654787, | |
| "eval_loss": 1.2012524604797363, | |
| "eval_macro_fpr": 0.036531607457408626, | |
| "eval_macro_sensitivity": 0.39029268516816945, | |
| "eval_macro_specificity": 0.9740627875145957, | |
| "eval_precision": 0.6007307052856502, | |
| "eval_precision_macro": 0.32788772117271575, | |
| "eval_recall": 0.6553059643687065, | |
| "eval_recall_macro": 0.39029268516816945, | |
| "eval_runtime": 30.3344, | |
| "eval_samples_per_second": 42.559, | |
| "eval_steps_per_second": 5.34, | |
| "eval_weighted_fpr": 0.03621124582960371, | |
| "eval_weighted_sensitivity": 0.6553059643687065, | |
| "eval_weighted_specificity": 0.9556358483502301, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7381874515879163, | |
| "eval_f1_macro": 0.4250111578260666, | |
| "eval_f1_micro": 0.7381874515879164, | |
| "eval_f1_weighted": 0.7236865886821344, | |
| "eval_loss": 0.8278939127922058, | |
| "eval_macro_fpr": 0.024763530819684316, | |
| "eval_macro_sensitivity": 0.46578072689552413, | |
| "eval_macro_specificity": 0.9806295017830671, | |
| "eval_precision": 0.7211273477007717, | |
| "eval_precision_macro": 0.40923635739063985, | |
| "eval_recall": 0.7381874515879163, | |
| "eval_recall_macro": 0.46578072689552413, | |
| "eval_runtime": 34.091, | |
| "eval_samples_per_second": 37.869, | |
| "eval_steps_per_second": 4.752, | |
| "eval_weighted_fpr": 0.024707602339181286, | |
| "eval_weighted_sensitivity": 0.7381874515879163, | |
| "eval_weighted_specificity": 0.9712550751580882, | |
| "step": 321 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.7544539116963594, | |
| "eval_f1_macro": 0.44305220236653187, | |
| "eval_f1_micro": 0.7544539116963594, | |
| "eval_f1_weighted": 0.7304902543811719, | |
| "eval_loss": 0.713018000125885, | |
| "eval_macro_fpr": 0.0233195243156939, | |
| "eval_macro_sensitivity": 0.4769897256873969, | |
| "eval_macro_specificity": 0.9816367201302573, | |
| "eval_precision": 0.7255238407846915, | |
| "eval_precision_macro": 0.4800443784048848, | |
| "eval_recall": 0.7544539116963594, | |
| "eval_recall_macro": 0.4769897256873969, | |
| "eval_runtime": 32.6151, | |
| "eval_samples_per_second": 39.583, | |
| "eval_steps_per_second": 4.967, | |
| "eval_weighted_fpr": 0.02271912850283093, | |
| "eval_weighted_sensitivity": 0.7544539116963594, | |
| "eval_weighted_specificity": 0.9700968902575, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 3.958333333333333e-05, | |
| "loss": 1.1985, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.7823392718822618, | |
| "eval_f1_macro": 0.5047859025633034, | |
| "eval_f1_micro": 0.7823392718822618, | |
| "eval_f1_weighted": 0.766027283346833, | |
| "eval_loss": 0.6922410726547241, | |
| "eval_macro_fpr": 0.02001637486004151, | |
| "eval_macro_sensitivity": 0.528327896933921, | |
| "eval_macro_specificity": 0.9837585704961105, | |
| "eval_precision": 0.7594430092765442, | |
| "eval_precision_macro": 0.5188076394331661, | |
| "eval_recall": 0.7823392718822618, | |
| "eval_recall_macro": 0.528327896933921, | |
| "eval_runtime": 32.5532, | |
| "eval_samples_per_second": 39.658, | |
| "eval_steps_per_second": 4.976, | |
| "eval_weighted_fpr": 0.019485472574717427, | |
| "eval_weighted_sensitivity": 0.7823392718822618, | |
| "eval_weighted_specificity": 0.9740392855593933, | |
| "step": 643 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.7939581719597212, | |
| "eval_f1_macro": 0.5257108179174788, | |
| "eval_f1_micro": 0.7939581719597211, | |
| "eval_f1_weighted": 0.7717932414170925, | |
| "eval_loss": 0.6710420846939087, | |
| "eval_macro_fpr": 0.018964580711785816, | |
| "eval_macro_sensitivity": 0.5571328425020213, | |
| "eval_macro_specificity": 0.9845265708178329, | |
| "eval_precision": 0.7734327517428056, | |
| "eval_precision_macro": 0.5449998556573111, | |
| "eval_recall": 0.7939581719597212, | |
| "eval_recall_macro": 0.5571328425020213, | |
| "eval_runtime": 33.235, | |
| "eval_samples_per_second": 38.845, | |
| "eval_steps_per_second": 4.874, | |
| "eval_weighted_fpr": 0.018199233716475097, | |
| "eval_weighted_sensitivity": 0.7939581719597212, | |
| "eval_weighted_specificity": 0.9739403903077732, | |
| "step": 803 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.7970565453137103, | |
| "eval_f1_macro": 0.5316254099721903, | |
| "eval_f1_micro": 0.7970565453137104, | |
| "eval_f1_weighted": 0.7789523196246046, | |
| "eval_loss": 0.6454855799674988, | |
| "eval_macro_fpr": 0.018368947118040532, | |
| "eval_macro_sensitivity": 0.5622105847462606, | |
| "eval_macro_specificity": 0.9848273693754934, | |
| "eval_precision": 0.7756987894676903, | |
| "eval_precision_macro": 0.5353415355904975, | |
| "eval_recall": 0.7970565453137103, | |
| "eval_recall_macro": 0.5622105847462606, | |
| "eval_runtime": 33.0425, | |
| "eval_samples_per_second": 39.071, | |
| "eval_steps_per_second": 4.903, | |
| "eval_weighted_fpr": 0.017862012544314153, | |
| "eval_weighted_sensitivity": 0.7970565453137103, | |
| "eval_weighted_specificity": 0.9753539953186898, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 2.916666666666667e-05, | |
| "loss": 0.5721, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8001549186676995, | |
| "eval_f1_macro": 0.5486067787470349, | |
| "eval_f1_micro": 0.8001549186676995, | |
| "eval_f1_weighted": 0.7845126172395842, | |
| "eval_loss": 0.6394562721252441, | |
| "eval_macro_fpr": 0.01807650265580631, | |
| "eval_macro_sensitivity": 0.5784319115133939, | |
| "eval_macro_specificity": 0.9850891561725063, | |
| "eval_precision": 0.7801018097427485, | |
| "eval_precision_macro": 0.5442815357986059, | |
| "eval_recall": 0.8001549186676995, | |
| "eval_recall_macro": 0.5784319115133939, | |
| "eval_runtime": 32.9727, | |
| "eval_samples_per_second": 39.154, | |
| "eval_steps_per_second": 4.913, | |
| "eval_weighted_fpr": 0.017527173913043478, | |
| "eval_weighted_sensitivity": 0.8001549186676995, | |
| "eval_weighted_specificity": 0.9761824239198935, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8024786986831913, | |
| "eval_f1_macro": 0.5474570084709011, | |
| "eval_f1_micro": 0.8024786986831914, | |
| "eval_f1_weighted": 0.787386582262622, | |
| "eval_loss": 0.6317066550254822, | |
| "eval_macro_fpr": 0.017822080148589308, | |
| "eval_macro_sensitivity": 0.5773035410379497, | |
| "eval_macro_specificity": 0.9852623885220758, | |
| "eval_precision": 0.783268466071204, | |
| "eval_precision_macro": 0.5438628735250125, | |
| "eval_recall": 0.8024786986831913, | |
| "eval_recall_macro": 0.5773035410379497, | |
| "eval_runtime": 33.0187, | |
| "eval_samples_per_second": 39.099, | |
| "eval_steps_per_second": 4.906, | |
| "eval_weighted_fpr": 0.017277593332881633, | |
| "eval_weighted_sensitivity": 0.8024786986831913, | |
| "eval_weighted_specificity": 0.9764571291479465, | |
| "step": 1286 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.8009295120061968, | |
| "eval_f1_macro": 0.5609368334662593, | |
| "eval_f1_micro": 0.8009295120061968, | |
| "eval_f1_weighted": 0.7875081013739925, | |
| "eval_loss": 0.6136514544487, | |
| "eval_macro_fpr": 0.017931645431496272, | |
| "eval_macro_sensitivity": 0.5841960959288544, | |
| "eval_macro_specificity": 0.985161694874728, | |
| "eval_precision": 0.7827917488346946, | |
| "eval_precision_macro": 0.5593054718395108, | |
| "eval_recall": 0.8009295120061968, | |
| "eval_recall_macro": 0.5841960959288544, | |
| "eval_runtime": 33.7216, | |
| "eval_samples_per_second": 38.284, | |
| "eval_steps_per_second": 4.804, | |
| "eval_weighted_fpr": 0.017443833570895267, | |
| "eval_weighted_sensitivity": 0.8009295120061968, | |
| "eval_weighted_specificity": 0.9764959111147221, | |
| "step": 1446 | |
| }, | |
| { | |
| "epoch": 9.33, | |
| "learning_rate": 1.8750000000000002e-05, | |
| "loss": 0.4166, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.8156467854376452, | |
| "eval_f1_macro": 0.6305053806449583, | |
| "eval_f1_micro": 0.8156467854376452, | |
| "eval_f1_weighted": 0.8066651282863864, | |
| "eval_loss": 0.6248635649681091, | |
| "eval_macro_fpr": 0.016463869736407642, | |
| "eval_macro_sensitivity": 0.6430050326236179, | |
| "eval_macro_specificity": 0.9861908495694477, | |
| "eval_precision": 0.8054630087428596, | |
| "eval_precision_macro": 0.639776904615918, | |
| "eval_recall": 0.8156467854376452, | |
| "eval_recall_macro": 0.6430050326236179, | |
| "eval_runtime": 33.5965, | |
| "eval_samples_per_second": 38.427, | |
| "eval_steps_per_second": 4.822, | |
| "eval_weighted_fpr": 0.01588785046728972, | |
| "eval_weighted_sensitivity": 0.8156467854376452, | |
| "eval_weighted_specificity": 0.9772159581040682, | |
| "step": 1607 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_accuracy": 0.8125484120836561, | |
| "eval_f1_macro": 0.6371983241862614, | |
| "eval_f1_micro": 0.8125484120836561, | |
| "eval_f1_weighted": 0.8042305582147614, | |
| "eval_loss": 0.642641544342041, | |
| "eval_macro_fpr": 0.016876700974845713, | |
| "eval_macro_sensitivity": 0.6520338532213917, | |
| "eval_macro_specificity": 0.9859193883690203, | |
| "eval_precision": 0.8013591068710365, | |
| "eval_precision_macro": 0.6397179331634546, | |
| "eval_recall": 0.8125484120836561, | |
| "eval_recall_macro": 0.6520338532213917, | |
| "eval_runtime": 33.85, | |
| "eval_samples_per_second": 38.139, | |
| "eval_steps_per_second": 4.786, | |
| "eval_weighted_fpr": 0.016211146838156484, | |
| "eval_weighted_sensitivity": 0.8125484120836561, | |
| "eval_weighted_specificity": 0.9762424134516496, | |
| "step": 1768 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_accuracy": 0.8164213787761425, | |
| "eval_f1_macro": 0.6372154277689757, | |
| "eval_f1_micro": 0.8164213787761425, | |
| "eval_f1_weighted": 0.8082672837501724, | |
| "eval_loss": 0.6305465698242188, | |
| "eval_macro_fpr": 0.016428781045237314, | |
| "eval_macro_sensitivity": 0.6526108250586471, | |
| "eval_macro_specificity": 0.9862261630603141, | |
| "eval_precision": 0.8050297561179627, | |
| "eval_precision_macro": 0.635785837139922, | |
| "eval_recall": 0.8164213787761425, | |
| "eval_recall_macro": 0.6526108250586471, | |
| "eval_runtime": 33.9194, | |
| "eval_samples_per_second": 38.061, | |
| "eval_steps_per_second": 4.776, | |
| "eval_weighted_fpr": 0.01580737677582872, | |
| "eval_weighted_sensitivity": 0.8164213787761425, | |
| "eval_weighted_specificity": 0.9769710671285662, | |
| "step": 1929 | |
| }, | |
| { | |
| "epoch": 12.44, | |
| "learning_rate": 8.333333333333334e-06, | |
| "loss": 0.3406, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_accuracy": 0.820294345468629, | |
| "eval_f1_macro": 0.6352955893706947, | |
| "eval_f1_micro": 0.8202943454686291, | |
| "eval_f1_weighted": 0.8129160642124462, | |
| "eval_loss": 0.6275700926780701, | |
| "eval_macro_fpr": 0.015989017879060468, | |
| "eval_macro_sensitivity": 0.6466831772856799, | |
| "eval_macro_specificity": 0.9865122165499488, | |
| "eval_precision": 0.8102004267117454, | |
| "eval_precision_macro": 0.6417861741800394, | |
| "eval_recall": 0.820294345468629, | |
| "eval_recall_macro": 0.6466831772856799, | |
| "eval_runtime": 32.8101, | |
| "eval_samples_per_second": 39.348, | |
| "eval_steps_per_second": 4.938, | |
| "eval_weighted_fpr": 0.015407092575375215, | |
| "eval_weighted_sensitivity": 0.820294345468629, | |
| "eval_weighted_specificity": 0.9773889027806014, | |
| "step": 2089 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_accuracy": 0.8187451587916343, | |
| "eval_f1_macro": 0.6446278757638214, | |
| "eval_f1_micro": 0.8187451587916342, | |
| "eval_f1_weighted": 0.81074292931159, | |
| "eval_loss": 0.6428066492080688, | |
| "eval_macro_fpr": 0.016186760254085688, | |
| "eval_macro_sensitivity": 0.6618493798882754, | |
| "eval_macro_specificity": 0.9863911143208972, | |
| "eval_precision": 0.8079041434352637, | |
| "eval_precision_macro": 0.6466770185879867, | |
| "eval_recall": 0.8187451587916343, | |
| "eval_recall_macro": 0.6618493798882754, | |
| "eval_runtime": 32.4757, | |
| "eval_samples_per_second": 39.753, | |
| "eval_steps_per_second": 4.988, | |
| "eval_weighted_fpr": 0.015566790846194785, | |
| "eval_weighted_sensitivity": 0.8187451587916343, | |
| "eval_weighted_specificity": 0.9771215560218227, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 14.93, | |
| "eval_accuracy": 0.820294345468629, | |
| "eval_f1_macro": 0.6460817297487148, | |
| "eval_f1_micro": 0.8202943454686291, | |
| "eval_f1_weighted": 0.8125066936495612, | |
| "eval_loss": 0.634365439414978, | |
| "eval_macro_fpr": 0.016027569664931556, | |
| "eval_macro_sensitivity": 0.6624718661648383, | |
| "eval_macro_specificity": 0.9864925668028308, | |
| "eval_precision": 0.8092174819516477, | |
| "eval_precision_macro": 0.6486629857671901, | |
| "eval_recall": 0.820294345468629, | |
| "eval_recall_macro": 0.6624718661648383, | |
| "eval_runtime": 32.8393, | |
| "eval_samples_per_second": 39.313, | |
| "eval_steps_per_second": 4.933, | |
| "eval_weighted_fpr": 0.015407092575375215, | |
| "eval_weighted_sensitivity": 0.820294345468629, | |
| "eval_weighted_specificity": 0.9770941565738336, | |
| "step": 2400 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 2400, | |
| "num_train_epochs": 15, | |
| "save_steps": 500, | |
| "total_flos": 2.082121538560819e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |