| { | |
| "best_metric": 0.7726753332748134, | |
| "best_model_checkpoint": "legal-bert-lora-no-grad/checkpoint-9002", | |
| "epoch": 18.0, | |
| "eval_steps": 500, | |
| "global_step": 11574, | |
| "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 | |
| }, | |
| { | |
| "epoch": 9.33, | |
| "learning_rate": 3.4447900466562985e-05, | |
| "loss": 0.195, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.8272656855151046, | |
| "eval_f1_macro": 0.752814241629395, | |
| "eval_f1_micro": 0.8272656855151046, | |
| "eval_f1_weighted": 0.8241119102181389, | |
| "eval_loss": 0.9685066342353821, | |
| "eval_macro_fpr": 0.01532247779991636, | |
| "eval_macro_sensitivity": 0.7514626154154147, | |
| "eval_macro_specificity": 0.9869341007498275, | |
| "eval_precision": 0.8235616888844944, | |
| "eval_precision_macro": 0.7594602286715586, | |
| "eval_recall": 0.8272656855151046, | |
| "eval_recall_macro": 0.7514626154154147, | |
| "eval_runtime": 32.9707, | |
| "eval_samples_per_second": 39.156, | |
| "eval_steps_per_second": 4.913, | |
| "eval_weighted_fpr": 0.014695222405271829, | |
| "eval_weighted_sensitivity": 0.8272656855151046, | |
| "eval_weighted_specificity": 0.9767458257323062, | |
| "step": 6430 | |
| }, | |
| { | |
| "epoch": 10.11, | |
| "learning_rate": 3.315189217210991e-05, | |
| "loss": 0.2071, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 10.89, | |
| "learning_rate": 3.1855883877656815e-05, | |
| "loss": 0.1617, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_accuracy": 0.831138652207591, | |
| "eval_f1_macro": 0.7542626782968901, | |
| "eval_f1_micro": 0.831138652207591, | |
| "eval_f1_weighted": 0.8265278944879543, | |
| "eval_loss": 1.0405933856964111, | |
| "eval_macro_fpr": 0.014948189971054987, | |
| "eval_macro_sensitivity": 0.7552299472419968, | |
| "eval_macro_specificity": 0.9872461533889039, | |
| "eval_precision": 0.8263112206278148, | |
| "eval_precision_macro": 0.7615032315409404, | |
| "eval_recall": 0.831138652207591, | |
| "eval_recall_macro": 0.7552299472419968, | |
| "eval_runtime": 32.4814, | |
| "eval_samples_per_second": 39.746, | |
| "eval_steps_per_second": 4.987, | |
| "eval_weighted_fpr": 0.014304461942257218, | |
| "eval_weighted_sensitivity": 0.831138652207591, | |
| "eval_weighted_specificity": 0.9775536486259658, | |
| "step": 7073 | |
| }, | |
| { | |
| "epoch": 11.66, | |
| "learning_rate": 3.0559875583203736e-05, | |
| "loss": 0.1421, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_accuracy": 0.8319132455460883, | |
| "eval_f1_macro": 0.7546443355775483, | |
| "eval_f1_micro": 0.8319132455460883, | |
| "eval_f1_weighted": 0.8286773098434994, | |
| "eval_loss": 1.0712851285934448, | |
| "eval_macro_fpr": 0.014849682505759526, | |
| "eval_macro_sensitivity": 0.7533324298517142, | |
| "eval_macro_specificity": 0.9872830280264108, | |
| "eval_precision": 0.8275778384582335, | |
| "eval_precision_macro": 0.7625814308958907, | |
| "eval_recall": 0.8319132455460883, | |
| "eval_recall_macro": 0.7533324298517142, | |
| "eval_runtime": 33.7221, | |
| "eval_samples_per_second": 38.284, | |
| "eval_steps_per_second": 4.804, | |
| "eval_weighted_fpr": 0.014226709499770536, | |
| "eval_weighted_sensitivity": 0.8319132455460883, | |
| "eval_weighted_specificity": 0.977332174850073, | |
| "step": 7716 | |
| }, | |
| { | |
| "epoch": 12.44, | |
| "learning_rate": 2.926386728875065e-05, | |
| "loss": 0.1184, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_accuracy": 0.82571649883811, | |
| "eval_f1_macro": 0.7510253218960331, | |
| "eval_f1_micro": 0.82571649883811, | |
| "eval_f1_weighted": 0.8218884223471629, | |
| "eval_loss": 1.1124904155731201, | |
| "eval_macro_fpr": 0.015517240035600281, | |
| "eval_macro_sensitivity": 0.7503900667536906, | |
| "eval_macro_specificity": 0.9868175907777439, | |
| "eval_precision": 0.8208602037612678, | |
| "eval_precision_macro": 0.7569190431250081, | |
| "eval_recall": 0.82571649883811, | |
| "eval_recall_macro": 0.7503900667536906, | |
| "eval_runtime": 32.7063, | |
| "eval_samples_per_second": 39.473, | |
| "eval_steps_per_second": 4.953, | |
| "eval_weighted_fpr": 0.01485246550927454, | |
| "eval_weighted_sensitivity": 0.82571649883811, | |
| "eval_weighted_specificity": 0.9765473628280475, | |
| "step": 8359 | |
| }, | |
| { | |
| "epoch": 13.22, | |
| "learning_rate": 2.7967858994297562e-05, | |
| "loss": 0.1157, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "learning_rate": 2.667185069984448e-05, | |
| "loss": 0.1017, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_accuracy": 0.8210689388071263, | |
| "eval_f1_macro": 0.7726753332748134, | |
| "eval_f1_micro": 0.8210689388071263, | |
| "eval_f1_weighted": 0.8196057150820856, | |
| "eval_loss": 1.1926428079605103, | |
| "eval_macro_fpr": 0.01592832331319769, | |
| "eval_macro_sensitivity": 0.7814792076835096, | |
| "eval_macro_specificity": 0.9865762698201747, | |
| "eval_precision": 0.8214722394566816, | |
| "eval_precision_macro": 0.7675145052350089, | |
| "eval_recall": 0.8210689388071263, | |
| "eval_recall_macro": 0.7814792076835096, | |
| "eval_runtime": 33.1567, | |
| "eval_samples_per_second": 38.936, | |
| "eval_steps_per_second": 4.886, | |
| "eval_weighted_fpr": 0.015327450069670227, | |
| "eval_weighted_sensitivity": 0.8210689388071263, | |
| "eval_weighted_specificity": 0.977575108495493, | |
| "step": 9002 | |
| }, | |
| { | |
| "epoch": 14.77, | |
| "learning_rate": 2.53758424053914e-05, | |
| "loss": 0.0752, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_accuracy": 0.8164213787761425, | |
| "eval_f1_macro": 0.740181771196266, | |
| "eval_f1_micro": 0.8164213787761425, | |
| "eval_f1_weighted": 0.8133417808004717, | |
| "eval_loss": 1.2507578134536743, | |
| "eval_macro_fpr": 0.016423646764491533, | |
| "eval_macro_sensitivity": 0.7376546806118065, | |
| "eval_macro_specificity": 0.9861141336738363, | |
| "eval_precision": 0.8120522617771967, | |
| "eval_precision_macro": 0.7479093161875279, | |
| "eval_recall": 0.8164213787761425, | |
| "eval_recall_macro": 0.7376546806118065, | |
| "eval_runtime": 33.1196, | |
| "eval_samples_per_second": 38.98, | |
| "eval_steps_per_second": 4.891, | |
| "eval_weighted_fpr": 0.01580737677582872, | |
| "eval_weighted_sensitivity": 0.8164213787761425, | |
| "eval_weighted_specificity": 0.9752906263314008, | |
| "step": 9645 | |
| }, | |
| { | |
| "epoch": 15.55, | |
| "learning_rate": 2.4079834110938313e-05, | |
| "loss": 0.0787, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_accuracy": 0.8218435321456236, | |
| "eval_f1_macro": 0.7698437404057171, | |
| "eval_f1_micro": 0.8218435321456236, | |
| "eval_f1_weighted": 0.8187688080779877, | |
| "eval_loss": 1.3246989250183105, | |
| "eval_macro_fpr": 0.015979873481610722, | |
| "eval_macro_sensitivity": 0.7584947789467064, | |
| "eval_macro_specificity": 0.9864726076618879, | |
| "eval_precision": 0.819938232209624, | |
| "eval_precision_macro": 0.8033763498836805, | |
| "eval_recall": 0.8218435321456236, | |
| "eval_recall_macro": 0.7584947789467064, | |
| "eval_runtime": 33.1643, | |
| "eval_samples_per_second": 38.927, | |
| "eval_steps_per_second": 4.885, | |
| "eval_weighted_fpr": 0.015247944842216919, | |
| "eval_weighted_sensitivity": 0.8218435321456236, | |
| "eval_weighted_specificity": 0.9752455827826969, | |
| "step": 10288 | |
| }, | |
| { | |
| "epoch": 16.33, | |
| "learning_rate": 2.2783825816485228e-05, | |
| "loss": 0.0668, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_accuracy": 0.8210689388071263, | |
| "eval_f1_macro": 0.7468315966002393, | |
| "eval_f1_micro": 0.8210689388071263, | |
| "eval_f1_weighted": 0.8197761848180135, | |
| "eval_loss": 1.3497052192687988, | |
| "eval_macro_fpr": 0.01580496987707635, | |
| "eval_macro_sensitivity": 0.7486969612108813, | |
| "eval_macro_specificity": 0.9865911342551769, | |
| "eval_precision": 0.8201285330679784, | |
| "eval_precision_macro": 0.7499841938739371, | |
| "eval_recall": 0.8210689388071263, | |
| "eval_recall_macro": 0.7486969612108813, | |
| "eval_runtime": 32.2442, | |
| "eval_samples_per_second": 40.038, | |
| "eval_steps_per_second": 5.024, | |
| "eval_weighted_fpr": 0.015327450069670227, | |
| "eval_weighted_sensitivity": 0.8210689388071263, | |
| "eval_weighted_specificity": 0.9777980750205295, | |
| "step": 10931 | |
| }, | |
| { | |
| "epoch": 17.11, | |
| "learning_rate": 2.1487817522032143e-05, | |
| "loss": 0.0457, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 17.88, | |
| "learning_rate": 2.0191809227579057e-05, | |
| "loss": 0.0471, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_accuracy": 0.8164213787761425, | |
| "eval_f1_macro": 0.7644188412597084, | |
| "eval_f1_micro": 0.8164213787761425, | |
| "eval_f1_weighted": 0.8159145099342531, | |
| "eval_loss": 1.4277746677398682, | |
| "eval_macro_fpr": 0.01645147998475456, | |
| "eval_macro_sensitivity": 0.7669665145653268, | |
| "eval_macro_specificity": 0.9861526001666732, | |
| "eval_precision": 0.8173561788436983, | |
| "eval_precision_macro": 0.7672032880393017, | |
| "eval_recall": 0.8164213787761425, | |
| "eval_recall_macro": 0.7669665145653268, | |
| "eval_runtime": 32.6421, | |
| "eval_samples_per_second": 39.55, | |
| "eval_steps_per_second": 4.963, | |
| "eval_weighted_fpr": 0.01580737677582872, | |
| "eval_weighted_sensitivity": 0.8164213787761425, | |
| "eval_weighted_specificity": 0.9758676237239569, | |
| "step": 11574 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 19290, | |
| "num_train_epochs": 30, | |
| "save_steps": 500, | |
| "total_flos": 2.510075191254221e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |