{"train/loss": 4.0074, "train/grad_norm": 0.39290162920951843, "train/learning_rate": 1.5833333333333336e-06, "train/epoch": 0.0, "train/global_step": 20, "_timestamp": 1713052645.0579474, "_runtime": 78.16842842102051, "_step": 0} |
{"train/loss": 4.0074, "train/grad_norm": 0.39290162920951843, "train/learning_rate": 1.5833333333333336e-06, "train/epoch": 0.0, "train/global_step": 20, "_timestamp": 1713052645.0579474, "_runtime": 78.16842842102051, "_step": 0} |