all_results.json

{"eval_train_loss": 0.04834188323334924, "eval_epoch": 49, "eval_step": 35000}