{"train/loss": 3.5148, "train/grad_norm": 27.329477310180664, "train/learning_rate": 2.90410829484944e-05, "train/epoch": 0.23, "train/global_step": 2820, "_timestamp": 1712911104.8626254, "_runtime": 3460.354907274246, "_step": 140} |
{"train/loss": 3.5148, "train/grad_norm": 27.329477310180664, "train/learning_rate": 2.90410829484944e-05, "train/epoch": 0.23, "train/global_step": 2820, "_timestamp": 1712911104.8626254, "_runtime": 3460.354907274246, "_step": 140} |