| { | |
| "best_metric": 0.90625, | |
| "best_model_checkpoint": "/nfs/production/literature/amina-mardiyyah/new_data/OT-Entity-Extraction-Pipeline/model_outputs/hf/Variant_V2/reinit_llrd/4K_no_reinit_classifier_llrd0.9/BiomedNLP_BiomedBERT_base/checkpoint-1408", | |
| "epoch": 7.0, | |
| "eval_steps": 500, | |
| "global_step": 2464, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 0.3181160092353821, | |
| "learning_rate": 9.886363636363637e-06, | |
| "loss": 0.227, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9924264890389397, | |
| "eval_f1": 0.5577264653641208, | |
| "eval_loss": 0.01651068590581417, | |
| "eval_precision": 0.575091575091575, | |
| "eval_recall": 0.5413793103448276, | |
| "eval_runtime": 4.458, | |
| "eval_samples_per_second": 270.525, | |
| "eval_steps_per_second": 17.048, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 1.8566864728927612, | |
| "learning_rate": 1.9886363636363638e-05, | |
| "loss": 0.0131, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9970116568195816, | |
| "eval_f1": 0.8357723577235773, | |
| "eval_loss": 0.009072142653167248, | |
| "eval_precision": 0.7907692307692308, | |
| "eval_recall": 0.8862068965517241, | |
| "eval_runtime": 4.5121, | |
| "eval_samples_per_second": 267.28, | |
| "eval_steps_per_second": 16.844, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.004470929503440857, | |
| "learning_rate": 1.8901515151515153e-05, | |
| "loss": 0.0053, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9978328808996966, | |
| "eval_f1": 0.8819444444444444, | |
| "eval_loss": 0.005974676925688982, | |
| "eval_precision": 0.8881118881118881, | |
| "eval_recall": 0.8758620689655172, | |
| "eval_runtime": 4.488, | |
| "eval_samples_per_second": 268.719, | |
| "eval_steps_per_second": 16.934, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 1.6404166221618652, | |
| "learning_rate": 1.779040404040404e-05, | |
| "loss": 0.0026, | |
| "step": 1408 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.9981750575997445, | |
| "eval_f1": 0.90625, | |
| "eval_loss": 0.007304870057851076, | |
| "eval_precision": 0.9125874125874126, | |
| "eval_recall": 0.9, | |
| "eval_runtime": 4.4508, | |
| "eval_samples_per_second": 270.961, | |
| "eval_steps_per_second": 17.075, | |
| "step": 1408 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.021774714812636375, | |
| "learning_rate": 1.667929292929293e-05, | |
| "loss": 0.0016, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.9981066222597349, | |
| "eval_f1": 0.8957264957264958, | |
| "eval_loss": 0.006704295519739389, | |
| "eval_precision": 0.888135593220339, | |
| "eval_recall": 0.903448275862069, | |
| "eval_runtime": 4.4921, | |
| "eval_samples_per_second": 268.47, | |
| "eval_steps_per_second": 16.918, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 0.004604855552315712, | |
| "learning_rate": 1.5568181818181822e-05, | |
| "loss": 0.001, | |
| "step": 2112 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.9982663047197573, | |
| "eval_f1": 0.9031141868512111, | |
| "eval_loss": 0.0071674492210149765, | |
| "eval_precision": 0.90625, | |
| "eval_recall": 0.9, | |
| "eval_runtime": 4.7306, | |
| "eval_samples_per_second": 254.936, | |
| "eval_steps_per_second": 16.066, | |
| "step": 2112 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 0.013881128281354904, | |
| "learning_rate": 1.4457070707070708e-05, | |
| "loss": 0.0007, | |
| "step": 2464 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.9980838104797317, | |
| "eval_f1": 0.9014084507042254, | |
| "eval_loss": 0.00853455625474453, | |
| "eval_precision": 0.920863309352518, | |
| "eval_recall": 0.8827586206896552, | |
| "eval_runtime": 4.4218, | |
| "eval_samples_per_second": 272.737, | |
| "eval_steps_per_second": 17.187, | |
| "step": 2464 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 7040, | |
| "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": 3 | |
| } | |
| }, | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 805331029976316.0, | |
| "train_batch_size": 16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |