Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.18/train_results.json

{
    "total_flos": 5.62815163329864e+19,
    "train_loss": 0.0,
    "train_runtime": 1304.047,
    "train_samples": 125770,
    "train_samples_per_second": 289.338,
    "train_steps_per_second": 18.084
}