{"train/loss": 3.5451, "train/grad_norm": 36.46931076049805, "train/learning_rate": 2.8505486295849884e-05, "train/epoch": 0.27, "train/global_step": 3380, "_timestamp": 1712911803.0306854, "_runtime": 4158.522967338562, "_step": 168}