xls-r-ta / trainer_state.json
jejomi's picture
End of training
d86a541
raw
history blame contribute delete
601 Bytes
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.000429331959471063,
"global_step": 10,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"step": 10,
"total_flos": 3795111991065600.0,
"train_loss": 9.808407592773438,
"train_runtime": 41.7034,
"train_samples_per_second": 0.48,
"train_steps_per_second": 0.24
}
],
"max_steps": 10,
"num_train_epochs": 1,
"total_flos": 3795111991065600.0,
"trial_name": null,
"trial_params": null
}