|
{ |
|
"best_metric": 0.8706581352833638, |
|
"best_model_checkpoint": "distilbert-hate_speech18\\run-6\\checkpoint-240", |
|
"epoch": 1.0, |
|
"eval_steps": 500, |
|
"global_step": 240, |
|
"is_hyper_param_search": true, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 4.4189080485362326e-06, |
|
"loss": 0.4732, |
|
"step": 240 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"eval_accuracy": 0.8706581352833638, |
|
"eval_loss": 0.45438072085380554, |
|
"eval_runtime": 8.0476, |
|
"eval_samples_per_second": 271.882, |
|
"eval_steps_per_second": 17.024, |
|
"step": 240 |
|
} |
|
], |
|
"logging_steps": 500, |
|
"max_steps": 1920, |
|
"num_input_tokens_seen": 0, |
|
"num_train_epochs": 8, |
|
"save_steps": 500, |
|
"total_flos": 163370211234720.0, |
|
"trial_name": null, |
|
"trial_params": { |
|
"alpha": 0.4787113643124785, |
|
"learning_rate": 5.050180626898551e-06, |
|
"num_train_epochs": 8, |
|
"per_device_eval_batch_size": 16, |
|
"per_device_train_batch_size": 32, |
|
"temperature": 27 |
|
} |
|
} |
|
|