Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.13/train_results.json

{
    "ewc_loss": -0.7421875,
    "ewc_loss_diag": 2.944469451904297e-05,
    "ewc_loss_parallel": -0.00775146484375,
    "total_flos": 5.62815163329864e+19,
    "train_loss": 0.35840998991934647,
    "train_runtime": 47115.5066,
    "train_samples": 125770,
    "train_samples_per_second": 8.008,
    "train_steps_per_second": 0.501
}