Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.06/all_results.json

{
    "ewc_loss": 0.07620732486248016,
    "ewc_loss_diag": 3.6716461181640625e-05,
    "ewc_loss_parallel": 0.0003958623274229467,
    "total_flos": 5.62815163329864e+19,
    "train_loss": 0.46694955879437966,
    "train_runtime": 50208.5497,
    "train_samples": 125770,
    "train_samples_per_second": 7.515,
    "train_steps_per_second": 0.47
}