{"train/loss": 3.5903, "train/grad_norm": 27.53672218322754, "train/learning_rate": 2.9058100575095156e-05, "train/epoch": 0.22, "train/global_step": 2800, "_timestamp": 1712911079.7639155, "_runtime": 3435.256197452545, "_step": 139} |
{"train/loss": 3.5903, "train/grad_norm": 27.53672218322754, "train/learning_rate": 2.9058100575095156e-05, "train/epoch": 0.22, "train/global_step": 2800, "_timestamp": 1712911079.7639155, "_runtime": 3435.256197452545, "_step": 139} |