|
{ |
|
"best_metric": 0.7086666666666667, |
|
"best_model_checkpoint": "Bert_tinybert-distilled/run-1/checkpoint-516", |
|
"epoch": 3.0, |
|
"eval_steps": 500, |
|
"global_step": 516, |
|
"is_hyper_param_search": true, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 1.0, |
|
"grad_norm": 0.10513057559728622, |
|
"learning_rate": 0.0004697676642472769, |
|
"loss": 0.0495, |
|
"step": 172 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"eval_AUC": 0.8095893333333333, |
|
"eval_BAC": 0.5, |
|
"eval_MCC": 0.0, |
|
"eval_PR_AUC": 0.7634102804324954, |
|
"eval_accuracy": 0.5, |
|
"eval_loss": 0.04859978333115578, |
|
"eval_runtime": 6.131, |
|
"eval_samples_per_second": 244.658, |
|
"eval_steps_per_second": 7.666, |
|
"step": 172 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"grad_norm": 0.061181988567113876, |
|
"learning_rate": 0.00031317844283151793, |
|
"loss": 0.0486, |
|
"step": 344 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"eval_AUC": 0.8169422222222222, |
|
"eval_BAC": 0.6066666666666667, |
|
"eval_MCC": 0.34181792192366234, |
|
"eval_PR_AUC": 0.7733898212292494, |
|
"eval_accuracy": 0.6066666666666667, |
|
"eval_loss": 0.04842817783355713, |
|
"eval_runtime": 6.2071, |
|
"eval_samples_per_second": 241.657, |
|
"eval_steps_per_second": 7.572, |
|
"step": 344 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"grad_norm": 0.035214055329561234, |
|
"learning_rate": 0.00015658922141575897, |
|
"loss": 0.0484, |
|
"step": 516 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"eval_AUC": 0.7751502222222222, |
|
"eval_BAC": 0.7086666666666667, |
|
"eval_MCC": 0.47074801855813936, |
|
"eval_PR_AUC": 0.718076635825553, |
|
"eval_accuracy": 0.7086666666666667, |
|
"eval_loss": 0.048104096204042435, |
|
"eval_runtime": 6.2731, |
|
"eval_samples_per_second": 239.115, |
|
"eval_steps_per_second": 7.492, |
|
"step": 516 |
|
} |
|
], |
|
"logging_steps": 500, |
|
"max_steps": 688, |
|
"num_input_tokens_seen": 0, |
|
"num_train_epochs": 4, |
|
"save_steps": 500, |
|
"stateful_callbacks": { |
|
"TrainerControl": { |
|
"args": { |
|
"should_epoch_stop": false, |
|
"should_evaluate": false, |
|
"should_log": false, |
|
"should_save": true, |
|
"should_training_stop": false |
|
}, |
|
"attributes": {} |
|
} |
|
}, |
|
"total_flos": 1351133190000.0, |
|
"train_batch_size": 32, |
|
"trial_name": null, |
|
"trial_params": { |
|
"alpha": 0.06190377563487581, |
|
"learning_rate": 0.0006263568856630359, |
|
"num_train_epochs": 4, |
|
"temperature": 15 |
|
} |
|
} |
|
|