Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.16/train_results.json

{
    "ewc_loss": -2128.0,
    "ewc_loss_diag": 0.0050048828125,
    "ewc_loss_parallel": -2.125,
    "total_flos": 5.62815163329864e+19,
    "train_loss": -653.4477224089757,
    "train_runtime": 47383.3285,
    "train_samples": 125770,
    "train_samples_per_second": 7.963,
    "train_steps_per_second": 0.498
}