Llama-3.1-8B-Instruct_SFT_sciencev00.18/train_results.json

{
    "total_flos": 2.0853122720147177e+19,
    "train_loss": 0.9531466799229548,
    "train_runtime": 21268.4838,
    "train_samples": 145693,
    "train_samples_per_second": 13.7,
    "train_steps_per_second": 0.856
}