microsoft / mup

maximal update parametrization (µP)

Home Page:https://arxiv.org/abs/2203.03466

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

missing os import in mup/examples/MLP/main.py ?

james-simon opened this issue · comments

Just thought I'd let you know that, when I ran python mup/examples/MLP/main.py --load_base_shapes ./mup/examples/MLP/width64.bsh, I got an error at the very bottom (when saving the .tsv) because os wasn't imported (losing a few hours of runtime!). Adding import os to the header fixed that. Maybe I was doing something wrong, but thought I'd flag it just in case

This looks like an oversight on our end. Thanks for flagging!!