{"final_train_loss": 0.8204041719436646, "best_val_loss": 1.45723557472229, "total_train_time": 102.29941725730896, "avg_inference_tokens_per_second": 615.1389556104078} |
{"final_train_loss": 0.8204041719436646, "best_val_loss": 1.45723557472229, "total_train_time": 102.29941725730896, "avg_inference_tokens_per_second": 615.1389556104078} |