{"train/loss": 3.7538, "train/grad_norm": 21.58758544921875, "train/learning_rate": 2.9e-05, "train/epoch": 0.05, "train/global_step": 580, "_timestamp": 1712908376.9997563, "_runtime": 732.4920382499695, "_step": 28} |
{"train/loss": 3.7538, "train/grad_norm": 21.58758544921875, "train/learning_rate": 2.9e-05, "train/epoch": 0.05, "train/global_step": 580, "_timestamp": 1712908376.9997563, "_runtime": 732.4920382499695, "_step": 28} |