Llama-3.1-8B-Instruct_SFT_mathfisher_v00.02_s43/train_results.json

{
    "ewc_loss": 2.968311309814453e-05,
    "total_flos": 5.628159003328302e+19,
    "train_loss": 0.4093159021503956,
    "train_runtime": 46764.448,
    "train_samples": 125770,
    "train_samples_per_second": 8.068,
    "train_steps_per_second": 0.504
}