Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.08/train_results.json

{
    "ewc_loss": 0.008638550527393818,
    "ewc_loss_diag": 0.0,
    "ewc_loss_parallel": 8.638550207251683e-05,
    "total_flos": 5.62815163329864e+19,
    "train_loss": 0.3824227017627763,
    "train_runtime": 45517.1597,
    "train_samples": 125770,
    "train_samples_per_second": 8.289,
    "train_steps_per_second": 0.518
}