Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.04/train_results.json

{
    "ewc_loss": 0.0849609375,
    "ewc_loss_diag": 3.457069396972656e-05,
    "ewc_loss_parallel": 0.0005035400390625,
    "total_flos": 5.62815163329864e+19,
    "train_loss": 0.4737072293779522,
    "train_runtime": 48395.8245,
    "train_samples": 125770,
    "train_samples_per_second": 7.796,
    "train_steps_per_second": 0.487
}