{"train/loss": 3.5043, "train/grad_norm": 23.667312622070312, "train/learning_rate": 2.8881318444640564e-05, "train/epoch": 0.24, "train/global_step": 3000, "_timestamp": 1712911322.7230785, "_runtime": 3678.215360403061, "_step": 149} |
{"train/loss": 3.5043, "train/grad_norm": 23.667312622070312, "train/learning_rate": 2.8881318444640564e-05, "train/epoch": 0.24, "train/global_step": 3000, "_timestamp": 1712911322.7230785, "_runtime": 3678.215360403061, "_step": 149} |