Llama-3.1-8B-Instruct_SFT_sciencev00.05/train_results.json

{
    "total_flos": 2.1067883493711675e+19,
    "train_loss": 0.9522557386704232,
    "train_runtime": 21515.8506,
    "train_samples": 145693,
    "train_samples_per_second": 13.543,
    "train_steps_per_second": 0.846
}