| { | |
| "best_metric": 0.7829107328933658, | |
| "best_model_checkpoint": "logs/indian_build_rr/roberta-base/seed_1/checkpoint-682", | |
| "epoch": 14.0, | |
| "eval_steps": 500, | |
| "global_step": 868, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.6509204584925321, | |
| "eval_loss": 1.1796680688858032, | |
| "eval_macro-f1": 0.2406374552281572, | |
| "eval_micro-f1": 0.6509204584925321, | |
| "eval_precision-macro": 0.36507985738726567, | |
| "eval_precision-micro": 0.6509204584925321, | |
| "eval_recall-macro": 0.24249037703341708, | |
| "eval_recall-micro": 0.6509204584925321, | |
| "eval_runtime": 1.5286, | |
| "eval_samples_per_second": 19.625, | |
| "eval_steps_per_second": 5.233, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7349774227162209, | |
| "eval_loss": 0.8353763222694397, | |
| "eval_macro-f1": 0.5254710956125371, | |
| "eval_micro-f1": 0.7349774227162209, | |
| "eval_precision-macro": 0.5349847227650002, | |
| "eval_precision-micro": 0.7349774227162209, | |
| "eval_recall-macro": 0.5291052629047126, | |
| "eval_recall-micro": 0.7349774227162209, | |
| "eval_runtime": 1.6436, | |
| "eval_samples_per_second": 18.252, | |
| "eval_steps_per_second": 4.867, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.7342827370614797, | |
| "eval_loss": 0.8058456182479858, | |
| "eval_macro-f1": 0.536617185045311, | |
| "eval_micro-f1": 0.7342827370614797, | |
| "eval_precision-macro": 0.5558680331273026, | |
| "eval_precision-micro": 0.7342827370614797, | |
| "eval_recall-macro": 0.5381948279596481, | |
| "eval_recall-micro": 0.7342827370614797, | |
| "eval_runtime": 1.5882, | |
| "eval_samples_per_second": 18.889, | |
| "eval_steps_per_second": 5.037, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.7502605071205279, | |
| "eval_loss": 0.7717716097831726, | |
| "eval_macro-f1": 0.530010703703156, | |
| "eval_micro-f1": 0.7502605071205279, | |
| "eval_precision-macro": 0.6245985387634561, | |
| "eval_precision-micro": 0.7502605071205279, | |
| "eval_recall-macro": 0.5200955397553431, | |
| "eval_recall-micro": 0.7502605071205279, | |
| "eval_runtime": 1.601, | |
| "eval_samples_per_second": 18.738, | |
| "eval_steps_per_second": 4.997, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.7641542202153525, | |
| "eval_loss": 0.7306948900222778, | |
| "eval_macro-f1": 0.5578515609115684, | |
| "eval_micro-f1": 0.7641542202153525, | |
| "eval_precision-macro": 0.5889876526435496, | |
| "eval_precision-micro": 0.7641542202153525, | |
| "eval_recall-macro": 0.5462553107739512, | |
| "eval_recall-micro": 0.7641542202153525, | |
| "eval_runtime": 2.1379, | |
| "eval_samples_per_second": 14.032, | |
| "eval_steps_per_second": 3.742, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.774574505036471, | |
| "eval_loss": 0.7098783254623413, | |
| "eval_macro-f1": 0.5481274858908393, | |
| "eval_micro-f1": 0.774574505036471, | |
| "eval_precision-macro": 0.6076438792386994, | |
| "eval_precision-micro": 0.774574505036471, | |
| "eval_recall-macro": 0.5431283891127849, | |
| "eval_recall-micro": 0.774574505036471, | |
| "eval_runtime": 2.1932, | |
| "eval_samples_per_second": 13.679, | |
| "eval_steps_per_second": 3.648, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.7811740187565127, | |
| "eval_loss": 0.7071970701217651, | |
| "eval_macro-f1": 0.5261426411307122, | |
| "eval_micro-f1": 0.7811740187565127, | |
| "eval_precision-macro": 0.6089513985670642, | |
| "eval_precision-micro": 0.7811740187565127, | |
| "eval_recall-macro": 0.5125569147899907, | |
| "eval_recall-micro": 0.7811740187565127, | |
| "eval_runtime": 1.6371, | |
| "eval_samples_per_second": 18.326, | |
| "eval_steps_per_second": 4.887, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.7825633900659952, | |
| "eval_loss": 0.6919089555740356, | |
| "eval_macro-f1": 0.5675875408188613, | |
| "eval_micro-f1": 0.7825633900659952, | |
| "eval_precision-macro": 0.6321251715307294, | |
| "eval_precision-micro": 0.7825633900659952, | |
| "eval_recall-macro": 0.5470775441802167, | |
| "eval_recall-micro": 0.7825633900659952, | |
| "eval_runtime": 1.6786, | |
| "eval_samples_per_second": 17.872, | |
| "eval_steps_per_second": 4.766, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 8.064516129032258, | |
| "grad_norm": 7.422909736633301, | |
| "learning_rate": 1.7951612903225806e-05, | |
| "loss": 0.8758, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.7735324765543592, | |
| "eval_loss": 0.7503196597099304, | |
| "eval_macro-f1": 0.569622502930968, | |
| "eval_micro-f1": 0.7735324765543592, | |
| "eval_precision-macro": 0.5665598749803964, | |
| "eval_precision-micro": 0.7735324765543592, | |
| "eval_recall-macro": 0.5818475586367124, | |
| "eval_recall-micro": 0.7735324765543592, | |
| "eval_runtime": 1.5468, | |
| "eval_samples_per_second": 19.395, | |
| "eval_steps_per_second": 5.172, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.7783952761375478, | |
| "eval_loss": 0.7511970400810242, | |
| "eval_macro-f1": 0.5755339015228546, | |
| "eval_micro-f1": 0.7783952761375478, | |
| "eval_precision-macro": 0.6053985952851118, | |
| "eval_precision-micro": 0.7783952761375478, | |
| "eval_recall-macro": 0.5655629578179421, | |
| "eval_recall-micro": 0.7783952761375478, | |
| "eval_runtime": 2.2548, | |
| "eval_samples_per_second": 13.305, | |
| "eval_steps_per_second": 3.548, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_accuracy": 0.7829107328933658, | |
| "eval_loss": 0.7655877470970154, | |
| "eval_macro-f1": 0.591328685794211, | |
| "eval_micro-f1": 0.7829107328933658, | |
| "eval_precision-macro": 0.6085589543807931, | |
| "eval_precision-micro": 0.7829107328933658, | |
| "eval_recall-macro": 0.5834711775606751, | |
| "eval_recall-micro": 0.7829107328933658, | |
| "eval_runtime": 2.2566, | |
| "eval_samples_per_second": 13.295, | |
| "eval_steps_per_second": 3.545, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_accuracy": 0.7738798193817298, | |
| "eval_loss": 0.786118745803833, | |
| "eval_macro-f1": 0.5843444583481733, | |
| "eval_micro-f1": 0.7738798193817298, | |
| "eval_precision-macro": 0.5971774078353586, | |
| "eval_precision-micro": 0.7738798193817298, | |
| "eval_recall-macro": 0.5885123710730829, | |
| "eval_recall-micro": 0.7738798193817298, | |
| "eval_runtime": 1.5545, | |
| "eval_samples_per_second": 19.299, | |
| "eval_steps_per_second": 5.146, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_accuracy": 0.7780479333101772, | |
| "eval_loss": 0.8238919377326965, | |
| "eval_macro-f1": 0.5701476938599402, | |
| "eval_micro-f1": 0.7780479333101772, | |
| "eval_precision-macro": 0.5975031172688886, | |
| "eval_precision-micro": 0.7780479333101772, | |
| "eval_recall-macro": 0.5748658781079373, | |
| "eval_recall-micro": 0.7780479333101772, | |
| "eval_runtime": 1.5795, | |
| "eval_samples_per_second": 18.993, | |
| "eval_steps_per_second": 5.065, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_accuracy": 0.7797846474470302, | |
| "eval_loss": 0.8271887302398682, | |
| "eval_macro-f1": 0.592619371017184, | |
| "eval_micro-f1": 0.7797846474470302, | |
| "eval_precision-macro": 0.6088825353073621, | |
| "eval_precision-micro": 0.7797846474470302, | |
| "eval_recall-macro": 0.5868004304340952, | |
| "eval_recall-micro": 0.7797846474470302, | |
| "eval_runtime": 2.326, | |
| "eval_samples_per_second": 12.898, | |
| "eval_steps_per_second": 3.439, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "step": 868, | |
| "total_flos": 6.747257278287053e+16, | |
| "train_loss": 0.6624447870913739, | |
| "train_runtime": 488.5119, | |
| "train_samples_per_second": 10.112, | |
| "train_steps_per_second": 2.538 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 1240, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 20, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "EarlyStoppingCallback": { | |
| "args": { | |
| "early_stopping_patience": 3, | |
| "early_stopping_threshold": 0.0 | |
| }, | |
| "attributes": { | |
| "early_stopping_patience_counter": 0 | |
| } | |
| }, | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 6.747257278287053e+16, | |
| "train_batch_size": 4, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |