|
{ |
|
"best_metric": 0.06852555274963379, |
|
"best_model_checkpoint": "/content/cyner_securebert/checkpoint-500", |
|
"epoch": 10.0, |
|
"eval_steps": 500, |
|
"global_step": 3520, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 1.42, |
|
"learning_rate": 1.715909090909091e-05, |
|
"loss": 0.1929, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 1.42, |
|
"eval_accuracy": 0.9836284321272774, |
|
"eval_f1": 0.7673568818514007, |
|
"eval_loss": 0.06852555274963379, |
|
"eval_precision": 0.7334109429569267, |
|
"eval_recall": 0.8045977011494253, |
|
"eval_runtime": 4.4947, |
|
"eval_samples_per_second": 180.658, |
|
"eval_steps_per_second": 22.693, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 2.84, |
|
"learning_rate": 1.431818181818182e-05, |
|
"loss": 0.048, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 2.84, |
|
"eval_accuracy": 0.9836797536566589, |
|
"eval_f1": 0.7992277992277992, |
|
"eval_loss": 0.07445864379405975, |
|
"eval_precision": 0.8054474708171206, |
|
"eval_recall": 0.7931034482758621, |
|
"eval_runtime": 3.5334, |
|
"eval_samples_per_second": 229.809, |
|
"eval_steps_per_second": 28.868, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 4.26, |
|
"learning_rate": 1.1477272727272729e-05, |
|
"loss": 0.0299, |
|
"step": 1500 |
|
}, |
|
{ |
|
"epoch": 4.26, |
|
"eval_accuracy": 0.9856812933025404, |
|
"eval_f1": 0.8204811844540407, |
|
"eval_loss": 0.07195162028074265, |
|
"eval_precision": 0.7935560859188544, |
|
"eval_recall": 0.8492975734355045, |
|
"eval_runtime": 3.5155, |
|
"eval_samples_per_second": 230.977, |
|
"eval_steps_per_second": 29.014, |
|
"step": 1500 |
|
}, |
|
{ |
|
"epoch": 5.68, |
|
"learning_rate": 8.636363636363637e-06, |
|
"loss": 0.0199, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 5.68, |
|
"eval_accuracy": 0.984757505773672, |
|
"eval_f1": 0.8185812931575643, |
|
"eval_loss": 0.08455660194158554, |
|
"eval_precision": 0.8049382716049382, |
|
"eval_recall": 0.8326947637292464, |
|
"eval_runtime": 4.0911, |
|
"eval_samples_per_second": 198.478, |
|
"eval_steps_per_second": 24.932, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 7.1, |
|
"learning_rate": 5.795454545454546e-06, |
|
"loss": 0.014, |
|
"step": 2500 |
|
}, |
|
{ |
|
"epoch": 7.1, |
|
"eval_accuracy": 0.9847061842442905, |
|
"eval_f1": 0.817283950617284, |
|
"eval_loss": 0.08779067546129227, |
|
"eval_precision": 0.7909199522102748, |
|
"eval_recall": 0.8454661558109834, |
|
"eval_runtime": 4.0935, |
|
"eval_samples_per_second": 198.365, |
|
"eval_steps_per_second": 24.918, |
|
"step": 2500 |
|
}, |
|
{ |
|
"epoch": 8.52, |
|
"learning_rate": 2.954545454545455e-06, |
|
"loss": 0.0098, |
|
"step": 3000 |
|
}, |
|
{ |
|
"epoch": 8.52, |
|
"eval_accuracy": 0.9845008981267642, |
|
"eval_f1": 0.8034825870646767, |
|
"eval_loss": 0.09072290360927582, |
|
"eval_precision": 0.7830303030303031, |
|
"eval_recall": 0.8250319284802043, |
|
"eval_runtime": 4.1105, |
|
"eval_samples_per_second": 197.544, |
|
"eval_steps_per_second": 24.815, |
|
"step": 3000 |
|
}, |
|
{ |
|
"epoch": 9.94, |
|
"learning_rate": 1.1363636363636364e-07, |
|
"loss": 0.0073, |
|
"step": 3500 |
|
}, |
|
{ |
|
"epoch": 9.94, |
|
"eval_accuracy": 0.9851680780087246, |
|
"eval_f1": 0.8119925046845722, |
|
"eval_loss": 0.09167523682117462, |
|
"eval_precision": 0.7946210268948656, |
|
"eval_recall": 0.8301404853128991, |
|
"eval_runtime": 3.5796, |
|
"eval_samples_per_second": 226.841, |
|
"eval_steps_per_second": 28.495, |
|
"step": 3500 |
|
}, |
|
{ |
|
"epoch": 10.0, |
|
"step": 3520, |
|
"total_flos": 1129405245494868.0, |
|
"train_loss": 0.04578042336824265, |
|
"train_runtime": 633.1793, |
|
"train_samples_per_second": 44.379, |
|
"train_steps_per_second": 5.559 |
|
} |
|
], |
|
"logging_steps": 500, |
|
"max_steps": 3520, |
|
"num_train_epochs": 10, |
|
"save_steps": 500, |
|
"total_flos": 1129405245494868.0, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|