Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.07/train_results.json

{
    "ewc_loss": 0.08365151286125183,
    "ewc_loss_diag": 3.790855407714844e-05,
    "ewc_loss_parallel": 0.00045809714356437325,
    "total_flos": 5.62815163329864e+19,
    "train_loss": 0.4756091437999748,
    "train_runtime": 47393.4997,
    "train_samples": 125770,
    "train_samples_per_second": 7.961,
    "train_steps_per_second": 0.498
}