{"train/loss": 3.4893, "train/grad_norm": 19.461843490600586, "train/learning_rate": 5.862880896264688e-07, "train/epoch": 0.26, "train/global_step": 18660, "_timestamp": 1712942872.3497894, "_runtime": 19680.75881934166, "_step": 932} |
{"train/loss": 3.4893, "train/grad_norm": 19.461843490600586, "train/learning_rate": 5.862880896264688e-07, "train/epoch": 0.26, "train/global_step": 18660, "_timestamp": 1712942872.3497894, "_runtime": 19680.75881934166, "_step": 932} |