search

lucidrains / audiolm-pytorch

Implementation of AudioLM, a SOTA Language Modeling Approach to Audio Generation out of Google Research, in Pytorch
MIT License
2.33k stars 249 forks source link

pin sklearn exactly to 0.24.0 #198

Closed LWprogramming closed 1 year ago

LWprogramming commented 1 year ago

in case minibatchkmeans unpickling ends up being an issue. This is the only place scikit-learn appears in the repo based on my grep.

lucidrains commented 1 year ago

nice! thank you! want to bump the version while you are at it?

LWprogramming commented 1 year ago

oh right, yes done

lucidrains commented 1 year ago

awesome!

rsxdalv commented 1 year ago

this causes scikit-learn to be built on the installer's machine, resulting in:

https://github.com/lucidrains/audiolm-pytorch/issues/182 https://github.com/rsxdalv/tts-generation-webui/issues/61

lucidrains commented 1 year ago

@LWprogramming do you know which version of scikit-learn the minibatchmeans stops working?

lucidrains commented 1 year ago

for reference, don't see any restrictions for the fairseq package https://github.com/facebookresearch/fairseq/blob/main/setup.py#L191

lucidrains commented 1 year ago

oh i see the error message now

ok, what i'm going to do is allow setup.py to still work, but just add an assert if hubert is being used