Llama-3.1-8B-Instruct_SFT_mathv00.02/train_results.json

{
    "total_flos": 4.051147348618982e+19,
    "train_loss": 0.36618162162283807,
    "train_runtime": 35311.5111,
    "train_samples": 125770,
    "train_samples_per_second": 10.685,
    "train_steps_per_second": 0.668
}