karpathy / llama2.c

Inference Llama 2 in one file of pure C

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

mfu calculation

pure-water opened this issue · comments

Hi as far as I know

MFU = 6N + 12LHPQ

It is a kind of heustric value not strictlly accurate. I was just wondering is it more precise on the big language model case than the small toy number?