{"train/loss": 3.4933, "train/grad_norm": 27.148874282836914, "train/learning_rate": 2.8309521707095835e-05, "train/epoch": 0.28, "train/global_step": 3560, "_timestamp": 1712912014.9304147, "_runtime": 4370.422696590424, "_step": 177} |
{"train/loss": 3.4933, "train/grad_norm": 27.148874282836914, "train/learning_rate": 2.8309521707095835e-05, "train/epoch": 0.28, "train/global_step": 3560, "_timestamp": 1712912014.9304147, "_runtime": 4370.422696590424, "_step": 177} |