{"train/loss": 0.0137, "train/grad_norm": 1012.9119262695312, "train/learning_rate": 7.305699481865285e-08, "train/epoch": 25.0, "train/global_step": 14475, "_timestamp": 1718325226.388219, "_runtime": 38783.36533117294, "_step": 144, "train_runtime": 38644.776, "train_samples_per_second": 1.498, "train_steps_per_second": 0.375, "total_flos": 0.0, "train_loss": 0.2025738267978856}