Packaging improvements - Githubissues

chaiNNer-org / spandrel

Spandrel gives your project support for various PyTorch architectures meant for AI Super-Resolution, restoration, and inpainting. Based on the model support implemented in chaiNNer.

MIT License

139 stars 12 forks source link

Packaging improvements #83

Closed akx closed 9 months ago

akx commented 9 months ago

requires-python doesn't need an upper limit; the package works with a nightly version of Torch on Python 3.12, but couldn't otherwise be installed without --ignore-requires-python.
Things configured in setup.cfg can be configured in pyproject.toml.
wheel and build shouldn't be PEP517 build requirements; similarly, build requirements shouldn't be in the build extra used by the release step, since python -m build . will install them in the isolated venv.

joeyballentine commented 9 months ago

nightly version of Torch

This is why there's an upper limit. It might work fine with the nightly version of torch, but until that's stable then we can't really say we support it. That being said, if @RunDevelopment agrees with you and thinks we should loosen that, then I'm fine with it.

joeyballentine commented 9 months ago

And, thanks for this contribution! It'll definitely be cleaner to not have that separate setup file + extra deps. I just didn't know how to properly set that stuff up