{"train/loss": 3.531, "train/grad_norm": 55.637813568115234, "train/learning_rate": 2.2971664586314055e-05, "train/epoch": 0.55, "train/global_step": 6840, "_timestamp": 1712916005.019154, "_runtime": 8360.511435985565, "_step": 341} |
{"train/loss": 3.531, "train/grad_norm": 55.637813568115234, "train/learning_rate": 2.2971664586314055e-05, "train/epoch": 0.55, "train/global_step": 6840, "_timestamp": 1712916005.019154, "_runtime": 8360.511435985565, "_step": 341} |