|
{
|
|
"best_metric": 0.45191457867622375,
|
|
"best_model_checkpoint": "Models/Access_to_Communications/checkpoint-75",
|
|
"epoch": 5.0,
|
|
"eval_steps": 500,
|
|
"global_step": 75,
|
|
"is_hyper_param_search": false,
|
|
"is_local_process_zero": true,
|
|
"is_world_process_zero": true,
|
|
"log_history": [
|
|
{
|
|
"epoch": 1.0,
|
|
"eval_AUC": 0.75,
|
|
"eval_accuracy": 0.42857142857142855,
|
|
"eval_f1": 0.2571428571428571,
|
|
"eval_loss": 0.707037091255188,
|
|
"eval_runtime": 0.0181,
|
|
"eval_samples_per_second": 386.811,
|
|
"eval_steps_per_second": 110.517,
|
|
"step": 15
|
|
},
|
|
{
|
|
"epoch": 2.0,
|
|
"eval_AUC": 0.75,
|
|
"eval_accuracy": 0.5714285714285714,
|
|
"eval_f1": 0.5142857142857143,
|
|
"eval_loss": 0.6698740124702454,
|
|
"eval_runtime": 0.0169,
|
|
"eval_samples_per_second": 414.241,
|
|
"eval_steps_per_second": 118.354,
|
|
"step": 30
|
|
},
|
|
{
|
|
"epoch": 3.0,
|
|
"eval_AUC": 0.75,
|
|
"eval_accuracy": 0.7142857142857143,
|
|
"eval_f1": 0.7142857142857143,
|
|
"eval_loss": 0.57427978515625,
|
|
"eval_runtime": 0.0168,
|
|
"eval_samples_per_second": 416.68,
|
|
"eval_steps_per_second": 119.052,
|
|
"step": 45
|
|
},
|
|
{
|
|
"epoch": 4.0,
|
|
"eval_AUC": 0.75,
|
|
"eval_accuracy": 0.5714285714285714,
|
|
"eval_f1": 0.5523809523809523,
|
|
"eval_loss": 0.5580509305000305,
|
|
"eval_runtime": 0.0166,
|
|
"eval_samples_per_second": 420.614,
|
|
"eval_steps_per_second": 120.175,
|
|
"step": 60
|
|
},
|
|
{
|
|
"epoch": 5.0,
|
|
"eval_AUC": 0.8333333333333334,
|
|
"eval_accuracy": 0.8571428571428571,
|
|
"eval_f1": 0.8571428571428571,
|
|
"eval_loss": 0.45191457867622375,
|
|
"eval_runtime": 0.0169,
|
|
"eval_samples_per_second": 414.118,
|
|
"eval_steps_per_second": 118.319,
|
|
"step": 75
|
|
}
|
|
],
|
|
"logging_steps": 500,
|
|
"max_steps": 300,
|
|
"num_input_tokens_seen": 0,
|
|
"num_train_epochs": 20,
|
|
"save_steps": 500,
|
|
"total_flos": 13340552728800.0,
|
|
"train_batch_size": 4,
|
|
"trial_name": null,
|
|
"trial_params": null
|
|
}
|
|
|