{"train/loss": 4.6848, "train/grad_norm": 86.82076263427734, "train/learning_rate": 9e-06, "train/epoch": 0.0, "train/global_step": 180, "_timestamp": 1712907433.3538334, "_runtime": 77.6049473285675, "_step": 8} |
{"train/loss": 4.6848, "train/grad_norm": 86.82076263427734, "train/learning_rate": 9e-06, "train/epoch": 0.0, "train/global_step": 180, "_timestamp": 1712907433.3538334, "_runtime": 77.6049473285675, "_step": 8} |