{"train/loss": 0.6012, "train/grad_norm": 3.3899552822113037, "train/learning_rate": 0.0002, "train/epoch": 1.99, "train/global_step": 2000, "_timestamp": 1710061377.8150547, "_runtime": 7953.062962770462, "_step": 1} |
{"train/loss": 0.6012, "train/grad_norm": 3.3899552822113037, "train/learning_rate": 0.0002, "train/epoch": 1.99, "train/global_step": 2000, "_timestamp": 1710061377.8150547, "_runtime": 7953.062962770462, "_step": 1} |