{"train/loss": 4.4426, "train/grad_norm": 115.61023712158203, "train/learning_rate": 6e-06, "train/epoch": 0.0, "train/global_step": 120, "_timestamp": 1712907413.591552, "_runtime": 57.842665910720825, "_step": 5} |
{"train/loss": 4.4426, "train/grad_norm": 115.61023712158203, "train/learning_rate": 6e-06, "train/epoch": 0.0, "train/global_step": 120, "_timestamp": 1712907413.591552, "_runtime": 57.842665910720825, "_step": 5} |