{"train/loss": 3.5348, "train/grad_norm": 22.156579971313477, "train/learning_rate": 6.954854106421715e-07, "train/epoch": 0.26, "train/global_step": 18540, "_timestamp": 1712942744.08272, "_runtime": 19552.491750001907, "_step": 926} |
{"train/loss": 3.5348, "train/grad_norm": 22.156579971313477, "train/learning_rate": 6.954854106421715e-07, "train/epoch": 0.26, "train/global_step": 18540, "_timestamp": 1712942744.08272, "_runtime": 19552.491750001907, "_step": 926} |