Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.10/train_results.json

{
    "total_flos": 5.62815163329864e+19,
    "train_loss": 0.0,
    "train_runtime": 852.4232,
    "train_samples": 125770,
    "train_samples_per_second": 442.632,
    "train_steps_per_second": 27.666
}