Test on Windows, Linux, MacOS Intel

[scott-huberty]

This makes sure we test our CI's on the major platforms (Linux, Windows, macOS). For macOS we are testing arm64 and Intel.

I suspect we may see a bug with installing pylossless on windows that users have been reporting.

[scott-huberty]

The Windows install bug fix was simply to explicitly specify the character encoding when reading the README.md (On windows I think the default encoding isn't UTF-8).

We should probably always explicitly specify the encoding when reading files anyways... But for installation I think an even better fix might be to switch from the requirements.txt / setup.py combo to the more modern pyproject.toml.

Once CI's are green (fingers crossed! 🤞 ), @christian-oreilly I'll merge tomorrow if I don't receive an approval or request for changes.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 73.03%. Comparing base (5f2c341) to head (c670dd1).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #169   +/-   ##
=======================================
  Coverage   73.03%   73.03%           
=======================================
  Files          22       22           
  Lines        1157     1157           
=======================================
  Hits          845      845           
  Misses        312      312           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[scott-huberty]

FYI we hit some warnings when testing our pipeline with the MacOS Intel VM. It was just a dependency conflict.

The issue is that PyTorch stopped supporting MacOS Intel machines upon release of 2.3 I believe. So torch 2.2.* gets installed on the MacOS Intel VM, which doesn't play with with Numpy 2!. So the patch is to downgrade to Numpy 1 On the MacOS Intel VM only.

But since they were just warnings anyways (no errors!), I think I might just add something to the documentation to hint to any other MacOS Intel users out there, that if they care about suppressing these warnings, they should be able to safely downgrade to Numpy 1.

Leaving this here because both I and @christian-oreilly have Intel Macs 🙂