Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.01/all_results.json

{
    "ewc_loss": 0.041259765625,
    "ewc_loss_parallel": 4.124641418457031e-05,
    "ewc_loss_perp": 0.0,
    "total_flos": 5.62815163329864e+19,
    "train_loss": 0.43322460087045367,
    "train_runtime": 46186.2128,
    "train_samples": 125770,
    "train_samples_per_second": 8.169,
    "train_steps_per_second": 0.511
}