|
{ |
|
"best_metric": 0.1801992505788803, |
|
"best_model_checkpoint": "outputs/bart-base-detox/checkpoint-405", |
|
"epoch": 4.9953917050691246, |
|
"global_step": 675, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 9e-05, |
|
"loss": 0.3171, |
|
"step": 135 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"eval_loss": 0.20179320871829987, |
|
"eval_runtime": 11.3072, |
|
"eval_samples_per_second": 274.16, |
|
"eval_steps_per_second": 34.314, |
|
"step": 135 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"learning_rate": 8e-05, |
|
"loss": 0.1852, |
|
"step": 270 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"eval_loss": 0.18728719651699066, |
|
"eval_runtime": 11.1563, |
|
"eval_samples_per_second": 277.869, |
|
"eval_steps_per_second": 34.778, |
|
"step": 270 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"learning_rate": 7e-05, |
|
"loss": 0.1559, |
|
"step": 405 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"eval_loss": 0.1801992505788803, |
|
"eval_runtime": 11.3199, |
|
"eval_samples_per_second": 273.853, |
|
"eval_steps_per_second": 34.276, |
|
"step": 405 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"learning_rate": 6e-05, |
|
"loss": 0.1356, |
|
"step": 540 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"eval_loss": 0.18711909651756287, |
|
"eval_runtime": 11.3935, |
|
"eval_samples_per_second": 272.085, |
|
"eval_steps_per_second": 34.054, |
|
"step": 540 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"learning_rate": 5e-05, |
|
"loss": 0.1151, |
|
"step": 675 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"eval_loss": 0.19263018667697906, |
|
"eval_runtime": 11.4175, |
|
"eval_samples_per_second": 271.514, |
|
"eval_steps_per_second": 33.983, |
|
"step": 675 |
|
} |
|
], |
|
"max_steps": 1350, |
|
"num_train_epochs": 10, |
|
"total_flos": 2689437955276800.0, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|