| { | |
| "best_metric": 0.7571266313428899, | |
| "best_model_checkpoint": "legal-bert-lora-no-grad/checkpoint-3215", | |
| "epoch": 9.0, | |
| "eval_steps": 500, | |
| "global_step": 5787, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.870399170554692e-05, | |
| "loss": 1.6412, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7513555383423702, | |
| "eval_f1_macro": 0.4277259717095209, | |
| "eval_f1_micro": 0.7513555383423703, | |
| "eval_f1_weighted": 0.7282568827191913, | |
| "eval_loss": 0.7924586534500122, | |
| "eval_macro_fpr": 0.023743972756733803, | |
| "eval_macro_sensitivity": 0.47068996339347813, | |
| "eval_macro_specificity": 0.9814182576870886, | |
| "eval_precision": 0.7190272898782211, | |
| "eval_precision_macro": 0.4122596954919311, | |
| "eval_recall": 0.7513555383423702, | |
| "eval_recall_macro": 0.47068996339347813, | |
| "eval_runtime": 31.9941, | |
| "eval_samples_per_second": 40.351, | |
| "eval_steps_per_second": 5.063, | |
| "eval_weighted_fpr": 0.023091863894683835, | |
| "eval_weighted_sensitivity": 0.7513555383423702, | |
| "eval_weighted_specificity": 0.9699183269639584, | |
| "step": 643 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 4.740798341109383e-05, | |
| "loss": 0.7481, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7900852052672347, | |
| "eval_f1_macro": 0.5997806825026601, | |
| "eval_f1_micro": 0.7900852052672347, | |
| "eval_f1_weighted": 0.7768800571423996, | |
| "eval_loss": 0.6771801710128784, | |
| "eval_macro_fpr": 0.019203116699768798, | |
| "eval_macro_sensitivity": 0.6251797015497511, | |
| "eval_macro_specificity": 0.9842820395671547, | |
| "eval_precision": 0.7726267649469656, | |
| "eval_precision_macro": 0.595773614705663, | |
| "eval_recall": 0.7900852052672347, | |
| "eval_recall_macro": 0.6251797015497511, | |
| "eval_runtime": 33.2196, | |
| "eval_samples_per_second": 38.863, | |
| "eval_steps_per_second": 4.877, | |
| "eval_weighted_fpr": 0.018624149542986735, | |
| "eval_weighted_sensitivity": 0.7900852052672347, | |
| "eval_weighted_specificity": 0.9741453882400846, | |
| "step": 1286 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 4.6111975116640746e-05, | |
| "loss": 0.6465, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8048024786986832, | |
| "eval_f1_macro": 0.610991978534836, | |
| "eval_f1_micro": 0.804802478698683, | |
| "eval_f1_weighted": 0.7903758160570191, | |
| "eval_loss": 0.64998859167099, | |
| "eval_macro_fpr": 0.017565196384739988, | |
| "eval_macro_sensitivity": 0.6414017118652243, | |
| "eval_macro_specificity": 0.9854108558122262, | |
| "eval_precision": 0.7930969709541013, | |
| "eval_precision_macro": 0.6215844916422869, | |
| "eval_recall": 0.8048024786986832, | |
| "eval_recall_macro": 0.6414017118652243, | |
| "eval_runtime": 33.2556, | |
| "eval_samples_per_second": 38.821, | |
| "eval_steps_per_second": 4.871, | |
| "eval_weighted_fpr": 0.017029328287606435, | |
| "eval_weighted_sensitivity": 0.8048024786986832, | |
| "eval_weighted_specificity": 0.9763603584847095, | |
| "step": 1929 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 4.481596682218767e-05, | |
| "loss": 0.5559, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 4.3519958527734576e-05, | |
| "loss": 0.4707, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8094500387296669, | |
| "eval_f1_macro": 0.6424924237429005, | |
| "eval_f1_micro": 0.8094500387296669, | |
| "eval_f1_weighted": 0.801786166531795, | |
| "eval_loss": 0.6704407930374146, | |
| "eval_macro_fpr": 0.017261054466023888, | |
| "eval_macro_sensitivity": 0.6689319587406968, | |
| "eval_macro_specificity": 0.9855970345792379, | |
| "eval_precision": 0.8007850899141215, | |
| "eval_precision_macro": 0.6321597811403302, | |
| "eval_recall": 0.8094500387296669, | |
| "eval_recall_macro": 0.6689319587406968, | |
| "eval_runtime": 32.7354, | |
| "eval_samples_per_second": 39.437, | |
| "eval_steps_per_second": 4.949, | |
| "eval_weighted_fpr": 0.016536703414896478, | |
| "eval_weighted_sensitivity": 0.8094500387296669, | |
| "eval_weighted_specificity": 0.9745054799589014, | |
| "step": 2572 | |
| }, | |
| { | |
| "epoch": 4.67, | |
| "learning_rate": 4.22239502332815e-05, | |
| "loss": 0.4021, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8280402788536019, | |
| "eval_f1_macro": 0.7571266313428899, | |
| "eval_f1_micro": 0.8280402788536019, | |
| "eval_f1_weighted": 0.8219004240794867, | |
| "eval_loss": 0.7320332527160645, | |
| "eval_macro_fpr": 0.015397351101601975, | |
| "eval_macro_sensitivity": 0.7573467308503988, | |
| "eval_macro_specificity": 0.9869691873059906, | |
| "eval_precision": 0.8268949272272527, | |
| "eval_precision_macro": 0.7782079900717894, | |
| "eval_recall": 0.8280402788536019, | |
| "eval_recall_macro": 0.7573467308503988, | |
| "eval_runtime": 33.3254, | |
| "eval_samples_per_second": 38.739, | |
| "eval_steps_per_second": 4.861, | |
| "eval_weighted_fpr": 0.014616802739004477, | |
| "eval_weighted_sensitivity": 0.8280402788536019, | |
| "eval_weighted_specificity": 0.9764975307362563, | |
| "step": 3215 | |
| }, | |
| { | |
| "epoch": 5.44, | |
| "learning_rate": 4.092794193882841e-05, | |
| "loss": 0.3627, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8241673121611154, | |
| "eval_f1_macro": 0.7374017625053243, | |
| "eval_f1_micro": 0.8241673121611154, | |
| "eval_f1_weighted": 0.8222929814764423, | |
| "eval_loss": 0.6892128586769104, | |
| "eval_macro_fpr": 0.015594676805191453, | |
| "eval_macro_sensitivity": 0.7364879265668598, | |
| "eval_macro_specificity": 0.9867341217946985, | |
| "eval_precision": 0.822653270980506, | |
| "eval_precision_macro": 0.7430958863475831, | |
| "eval_recall": 0.8241673121611154, | |
| "eval_recall_macro": 0.7364879265668598, | |
| "eval_runtime": 32.5839, | |
| "eval_samples_per_second": 39.621, | |
| "eval_steps_per_second": 4.972, | |
| "eval_weighted_fpr": 0.015010249289162203, | |
| "eval_weighted_sensitivity": 0.8241673121611154, | |
| "eval_weighted_specificity": 0.9768445147593636, | |
| "step": 3858 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 3.963193364437533e-05, | |
| "loss": 0.3623, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 3.833592534992224e-05, | |
| "loss": 0.2866, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8179705654531371, | |
| "eval_f1_macro": 0.7443559250736709, | |
| "eval_f1_micro": 0.817970565453137, | |
| "eval_f1_weighted": 0.8121793367264196, | |
| "eval_loss": 0.8756108283996582, | |
| "eval_macro_fpr": 0.016562044787314777, | |
| "eval_macro_sensitivity": 0.7409975687139162, | |
| "eval_macro_specificity": 0.9859854959008175, | |
| "eval_precision": 0.8170676985045833, | |
| "eval_precision_macro": 0.7748072523920921, | |
| "eval_recall": 0.8179705654531371, | |
| "eval_recall_macro": 0.7409975687139162, | |
| "eval_runtime": 34.5334, | |
| "eval_samples_per_second": 37.384, | |
| "eval_steps_per_second": 4.691, | |
| "eval_weighted_fpr": 0.01564684732671949, | |
| "eval_weighted_sensitivity": 0.8179705654531371, | |
| "eval_weighted_specificity": 0.9718118730591274, | |
| "step": 4501 | |
| }, | |
| { | |
| "epoch": 7.78, | |
| "learning_rate": 3.7039917055469156e-05, | |
| "loss": 0.2639, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8264910921766073, | |
| "eval_f1_macro": 0.7480006271740781, | |
| "eval_f1_micro": 0.8264910921766073, | |
| "eval_f1_weighted": 0.8217020757246023, | |
| "eval_loss": 0.8579551577568054, | |
| "eval_macro_fpr": 0.0155160293084745, | |
| "eval_macro_sensitivity": 0.7427592056873393, | |
| "eval_macro_specificity": 0.9868042528312828, | |
| "eval_precision": 0.8259376217895409, | |
| "eval_precision_macro": 0.798909899893037, | |
| "eval_recall": 0.8264910921766073, | |
| "eval_recall_macro": 0.7427592056873393, | |
| "eval_runtime": 33.4037, | |
| "eval_samples_per_second": 38.648, | |
| "eval_steps_per_second": 4.85, | |
| "eval_weighted_fpr": 0.014773776546629732, | |
| "eval_weighted_sensitivity": 0.8264910921766073, | |
| "eval_weighted_specificity": 0.9755727002926348, | |
| "step": 5144 | |
| }, | |
| { | |
| "epoch": 8.55, | |
| "learning_rate": 3.574390876101607e-05, | |
| "loss": 0.2295, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.82571649883811, | |
| "eval_f1_macro": 0.7524178623807065, | |
| "eval_f1_micro": 0.82571649883811, | |
| "eval_f1_weighted": 0.822346219811999, | |
| "eval_loss": 0.9366260170936584, | |
| "eval_macro_fpr": 0.015511369010760676, | |
| "eval_macro_sensitivity": 0.7465079310173777, | |
| "eval_macro_specificity": 0.9867928280373104, | |
| "eval_precision": 0.8230871400583257, | |
| "eval_precision_macro": 0.7724676433487055, | |
| "eval_recall": 0.82571649883811, | |
| "eval_recall_macro": 0.7465079310173777, | |
| "eval_runtime": 33.4584, | |
| "eval_samples_per_second": 38.585, | |
| "eval_steps_per_second": 4.842, | |
| "eval_weighted_fpr": 0.01485246550927454, | |
| "eval_weighted_sensitivity": 0.82571649883811, | |
| "eval_weighted_specificity": 0.976175921721546, | |
| "step": 5787 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 19290, | |
| "num_train_epochs": 30, | |
| "save_steps": 500, | |
| "total_flos": 1.2550375956271104e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |