| { | |
| "best_metric": 0.7926363320597429, | |
| "best_model_checkpoint": "logs/indian_build_rr/MHGanainy/xmod-shared-roberta-base-legal-multi/seed_1/checkpoint-992", | |
| "epoch": 11.0, | |
| "eval_steps": 500, | |
| "global_step": 1364, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7460923931920805, | |
| "eval_loss": 0.8663367033004761, | |
| "eval_macro-f1": 0.47795631478138917, | |
| "eval_micro-f1": 0.7460923931920805, | |
| "eval_precision-macro": 0.537033221622001, | |
| "eval_precision-micro": 0.7460923931920805, | |
| "eval_recall-macro": 0.46833031643592427, | |
| "eval_recall-micro": 0.7460923931920805, | |
| "eval_runtime": 4.1397, | |
| "eval_samples_per_second": 7.247, | |
| "eval_steps_per_second": 3.623, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.6981590830149358, | |
| "eval_loss": 0.9012688398361206, | |
| "eval_macro-f1": 0.5023050615982335, | |
| "eval_micro-f1": 0.6981590830149358, | |
| "eval_precision-macro": 0.49382932128709284, | |
| "eval_precision-micro": 0.6981590830149358, | |
| "eval_recall-macro": 0.5409271662176228, | |
| "eval_recall-micro": 0.6981590830149358, | |
| "eval_runtime": 4.0156, | |
| "eval_samples_per_second": 7.471, | |
| "eval_steps_per_second": 3.735, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.7634595345606113, | |
| "eval_loss": 0.7796859741210938, | |
| "eval_macro-f1": 0.5415425168075806, | |
| "eval_micro-f1": 0.7634595345606113, | |
| "eval_precision-macro": 0.576524112938428, | |
| "eval_precision-micro": 0.7634595345606113, | |
| "eval_recall-macro": 0.5450809655959412, | |
| "eval_recall-micro": 0.7634595345606113, | |
| "eval_runtime": 4.0974, | |
| "eval_samples_per_second": 7.322, | |
| "eval_steps_per_second": 3.661, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.771795762417506, | |
| "eval_loss": 0.7202845215797424, | |
| "eval_macro-f1": 0.5408809683499423, | |
| "eval_micro-f1": 0.771795762417506, | |
| "eval_precision-macro": 0.6529799444819455, | |
| "eval_precision-micro": 0.771795762417506, | |
| "eval_recall-macro": 0.547754417411679, | |
| "eval_recall-micro": 0.771795762417506, | |
| "eval_runtime": 4.0517, | |
| "eval_samples_per_second": 7.404, | |
| "eval_steps_per_second": 3.702, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 4.032258064516129, | |
| "grad_norm": 6.1605224609375, | |
| "learning_rate": 2.398790322580645e-05, | |
| "loss": 0.9675, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.7777005904828065, | |
| "eval_loss": 0.7465451955795288, | |
| "eval_macro-f1": 0.5866418390333193, | |
| "eval_micro-f1": 0.7777005904828065, | |
| "eval_precision-macro": 0.5984466979286676, | |
| "eval_precision-micro": 0.7777005904828065, | |
| "eval_recall-macro": 0.5960188099213857, | |
| "eval_recall-micro": 0.7777005904828065, | |
| "eval_runtime": 4.1357, | |
| "eval_samples_per_second": 7.254, | |
| "eval_steps_per_second": 3.627, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.779089961792289, | |
| "eval_loss": 0.7502714991569519, | |
| "eval_macro-f1": 0.5699431692507972, | |
| "eval_micro-f1": 0.779089961792289, | |
| "eval_precision-macro": 0.613407043792223, | |
| "eval_precision-micro": 0.779089961792289, | |
| "eval_recall-macro": 0.5692211367836036, | |
| "eval_recall-micro": 0.779089961792289, | |
| "eval_runtime": 3.9786, | |
| "eval_samples_per_second": 7.54, | |
| "eval_steps_per_second": 3.77, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.7863841611670719, | |
| "eval_loss": 0.766509473323822, | |
| "eval_macro-f1": 0.5877444821246216, | |
| "eval_micro-f1": 0.7863841611670719, | |
| "eval_precision-macro": 0.6552426757086595, | |
| "eval_precision-micro": 0.7863841611670719, | |
| "eval_recall-macro": 0.5731587593411557, | |
| "eval_recall-micro": 0.7863841611670719, | |
| "eval_runtime": 4.3188, | |
| "eval_samples_per_second": 6.946, | |
| "eval_steps_per_second": 3.473, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.7926363320597429, | |
| "eval_loss": 0.765051007270813, | |
| "eval_macro-f1": 0.5936857133353053, | |
| "eval_micro-f1": 0.7926363320597429, | |
| "eval_precision-macro": 0.6253302610595257, | |
| "eval_precision-micro": 0.7926363320597429, | |
| "eval_recall-macro": 0.5880039852948196, | |
| "eval_recall-micro": 0.7926363320597429, | |
| "eval_runtime": 4.1192, | |
| "eval_samples_per_second": 7.283, | |
| "eval_steps_per_second": 3.641, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 8.064516129032258, | |
| "grad_norm": 4.945093154907227, | |
| "learning_rate": 1.793951612903226e-05, | |
| "loss": 0.5065, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.7766585620006947, | |
| "eval_loss": 0.8559873104095459, | |
| "eval_macro-f1": 0.594537265696571, | |
| "eval_micro-f1": 0.7766585620006947, | |
| "eval_precision-macro": 0.6074559744964598, | |
| "eval_precision-micro": 0.7766585620006947, | |
| "eval_recall-macro": 0.5930428516685216, | |
| "eval_recall-micro": 0.7766585620006947, | |
| "eval_runtime": 4.1612, | |
| "eval_samples_per_second": 7.209, | |
| "eval_steps_per_second": 3.605, | |
| "step": 1116 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.7902049322681487, | |
| "eval_loss": 0.8643214106559753, | |
| "eval_macro-f1": 0.5971614075338256, | |
| "eval_micro-f1": 0.7902049322681487, | |
| "eval_precision-macro": 0.6353648381926511, | |
| "eval_precision-micro": 0.7902049322681487, | |
| "eval_recall-macro": 0.5841529326694309, | |
| "eval_recall-micro": 0.7902049322681487, | |
| "eval_runtime": 4.3374, | |
| "eval_samples_per_second": 6.917, | |
| "eval_steps_per_second": 3.458, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_accuracy": 0.7895102466134074, | |
| "eval_loss": 0.9258854985237122, | |
| "eval_macro-f1": 0.6020711019594975, | |
| "eval_micro-f1": 0.7895102466134074, | |
| "eval_precision-macro": 0.6337053537013936, | |
| "eval_precision-micro": 0.7895102466134074, | |
| "eval_recall-macro": 0.5883671234037948, | |
| "eval_recall-micro": 0.7895102466134074, | |
| "eval_runtime": 3.9857, | |
| "eval_samples_per_second": 7.527, | |
| "eval_steps_per_second": 3.763, | |
| "step": 1364 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "step": 1364, | |
| "total_flos": 5.68024429585367e+16, | |
| "train_loss": 0.6171089015748145, | |
| "train_runtime": 1002.5092, | |
| "train_samples_per_second": 4.928, | |
| "train_steps_per_second": 2.474 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 2480, | |
| "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": 5.68024429585367e+16, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |