| {"train/loss": 2.9453, "train/grad_norm": 5.314031600952148, "train/learning_rate": 2.7450980392156865e-05, "train/epoch": 0.0, "train/global_step": 280, "_timestamp": 1712607940.627796, "_runtime": 463.31381702423096, "_step": 13, "_wandb": {"runtime": 480}} |