import nonechucks fails for torch1.4.0+cu100

[Hvitgar]

Importing nonechucks with torch1.4.0+cu100 (and possibly other torch versions with a cuda suffix) results in ValueError: invalid literal for int() with base 10: '0+cu100'

[sooheon]

Also fails for torch nightly, because of the torch version checking code in init.py

[XinDongol]

so issue

[waynchi]

Bump. Just got this same error on trying to import nonechucks in Google Colab.

Any temporary fix and any permanent fix planned?

Edit: It seems like there is a temporary fix here, #29