Llama-3.1-8B-Instruct_SFT_sciencev00.12/train_results.json

{
    "total_flos": 6.401115957451817e+17,
    "train_loss": 0.3628057053238609,
    "train_runtime": 1540.8029,
    "train_samples": 15696,
    "train_samples_per_second": 20.374,
    "train_steps_per_second": 1.273
}