{"train/loss": 7.2435, "train/grad_norm": 11.764897346496582, "train/learning_rate": 4.902572096648481e-05, "train/epoch": 0.39, "train/global_step": 1000, "_timestamp": 1717360413.9039454, "_runtime": 1180.4432513713837, "_step": 19}