|
{ |
|
"best_metric": 0.8983050584793091, |
|
"best_model_checkpoint": "OUTPUT/harish/PT-FalseTrue-0_2_BEST/SHORT/FalseTrue-0/2/checkpoint-14", |
|
"epoch": 7.0, |
|
"global_step": 14, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 1.0, |
|
"eval_accuracy": 0.7923728823661804, |
|
"eval_loss": 0.9869656562805176, |
|
"eval_runtime": 3.3495, |
|
"eval_samples_per_second": 70.457, |
|
"step": 2 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"eval_accuracy": 0.8474576473236084, |
|
"eval_loss": 0.5227280855178833, |
|
"eval_runtime": 3.3651, |
|
"eval_samples_per_second": 70.132, |
|
"step": 4 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"eval_accuracy": 0.8644067645072937, |
|
"eval_loss": 0.4343617260456085, |
|
"eval_runtime": 3.3829, |
|
"eval_samples_per_second": 69.762, |
|
"step": 6 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"eval_accuracy": 0.8728813529014587, |
|
"eval_loss": 0.3969828486442566, |
|
"eval_runtime": 3.514, |
|
"eval_samples_per_second": 67.16, |
|
"step": 8 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"eval_accuracy": 0.8771186470985413, |
|
"eval_loss": 0.3839848041534424, |
|
"eval_runtime": 3.3686, |
|
"eval_samples_per_second": 70.058, |
|
"step": 10 |
|
}, |
|
{ |
|
"epoch": 6.0, |
|
"eval_accuracy": 0.8898305296897888, |
|
"eval_loss": 0.33938831090927124, |
|
"eval_runtime": 3.4593, |
|
"eval_samples_per_second": 68.222, |
|
"step": 12 |
|
}, |
|
{ |
|
"epoch": 7.0, |
|
"eval_accuracy": 0.8983050584793091, |
|
"eval_loss": 0.31968358159065247, |
|
"eval_runtime": 3.3956, |
|
"eval_samples_per_second": 69.501, |
|
"step": 14 |
|
} |
|
], |
|
"max_steps": 18, |
|
"num_train_epochs": 9, |
|
"total_flos": 7785779516928.0, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|