Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.09/train_results.json

{
    "ewc_loss": 0.03533089905977249,
    "ewc_loss_diag": 0.0,
    "ewc_loss_parallel": 3.5330900573171675e-05,
    "total_flos": 5.62815163329864e+19,
    "train_loss": 0.42502481960762,
    "train_runtime": 45533.7805,
    "train_samples": 125770,
    "train_samples_per_second": 8.286,
    "train_steps_per_second": 0.518
}