{"train/loss": 3.9827, "train/grad_norm": 28.68460464477539, "train/learning_rate": 1.9e-05, "train/epoch": 0.03, "train/global_step": 380, "_timestamp": 1712908129.7701256, "_runtime": 485.262407541275, "_step": 18} |
{"train/loss": 3.9827, "train/grad_norm": 28.68460464477539, "train/learning_rate": 1.9e-05, "train/epoch": 0.03, "train/global_step": 380, "_timestamp": 1712908129.7701256, "_runtime": 485.262407541275, "_step": 18} |