{"final_train_loss": 0.9981321692466736, "best_val_loss": 0.9795330762863159, "total_train_time": 1213.3201868534088, "avg_inference_tokens_per_second": 575.7515195836003} |
{"final_train_loss": 0.9981321692466736, "best_val_loss": 0.9795330762863159, "total_train_time": 1213.3201868534088, "avg_inference_tokens_per_second": 575.7515195836003} |