|
{ |
|
"best_metric": 0.9864761697022576, |
|
"best_model_checkpoint": "final_models/glue_models/structroberta_s2_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": 0.98629230260849, |
|
"eval_f1": 0.9864761697022576, |
|
"eval_loss": 0.04553624615073204, |
|
"eval_mcc": 0.9729443961241362, |
|
"eval_runtime": 35.2221, |
|
"eval_samples_per_second": 513.655, |
|
"eval_steps_per_second": 64.221, |
|
"step": 400 |
|
}, |
|
{ |
|
"epoch": 6.58, |
|
"learning_rate": 1.7105263157894737e-05, |
|
"loss": 0.0135, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 10.0, |
|
"step": 760, |
|
"total_flos": 7567159074877440.0, |
|
"train_loss": 0.008936366656991213, |
|
"train_runtime": 427.2516, |
|
"train_samples_per_second": 212.24, |
|
"train_steps_per_second": 1.779 |
|
} |
|
], |
|
"max_steps": 760, |
|
"num_train_epochs": 10, |
|
"total_flos": 7567159074877440.0, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|