|
{ |
|
"best_metric": 0.9623766919018124, |
|
"best_model_checkpoint": "final_models/glue_models/structroberta_s1_final//finetune/relative_position_control/checkpoint-400", |
|
"epoch": 10.0, |
|
"global_step": 760, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 5.26, |
|
"eval_accuracy": 0.963740885257721, |
|
"eval_f1": 0.9623766919018124, |
|
"eval_loss": 0.16143889725208282, |
|
"eval_mcc": 0.9299301858120512, |
|
"eval_runtime": 35.4988, |
|
"eval_samples_per_second": 509.65, |
|
"eval_steps_per_second": 63.72, |
|
"step": 400 |
|
}, |
|
{ |
|
"epoch": 6.58, |
|
"learning_rate": 1.7105263157894737e-05, |
|
"loss": 0.0107, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 10.0, |
|
"step": 760, |
|
"total_flos": 7567165760532480.0, |
|
"train_loss": 0.007054869188485961, |
|
"train_runtime": 457.2791, |
|
"train_samples_per_second": 198.303, |
|
"train_steps_per_second": 1.662 |
|
} |
|
], |
|
"max_steps": 760, |
|
"num_train_epochs": 10, |
|
"total_flos": 7567165760532480.0, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|