| {"train/loss": 3.3771, "train/grad_norm": 6.049938201904297, "train/learning_rate": 3.5294117647058825e-06, "train/epoch": 0.0, "train/global_step": 120, "_timestamp": 1712608200.1904802, "_runtime": 217.42743515968323, "_step": 5, "_wandb": {"runtime": 237}} |