Llama-3.1-8B-Instruct_SFT_sciencev00.01/train_results.json

{
    "total_flos": 3.1829322264618205e+19,
    "train_loss": 0.922418457405818,
    "train_runtime": 33639.8393,
    "train_samples": 145693,
    "train_samples_per_second": 12.993,
    "train_steps_per_second": 0.812
}