hmm looks like macos-14 doesn't like it @ntkathole can you provide some guidance here?

i'm not sure i created my venv with make install-python-dependencies-dev

i'll start over with a venv created that way

looks like nvidia-* causes macos to fail and nvidia-* comes from torch / image comes from ci

looks like nvidia-* causes macos to fail and nvidia-* comes from torch / image comes from ci

The requirement file included in PR has nvidia-* dependencies which is only meant for GPU supported systems. You need to make sure make lock-python-dependencies-all do not include those dependencies, I think probably using PIP_EXTRA_INDEX_URL=https://download.pytorch.org/whl/cpu

Also, fyi - we have APACHE_ARROW_VERSION="17.0.0" added in feast-operator images, which needs to be updated as well.
https://github.com/search?q=repo%3Afeast-dev%2Ffeast%20APACHE_ARROW_VERSION&type=code

@ntkathole FYI: i've modified lock-python-dependencies-all to pass the args necessary to use the pytorch cpu index

its not clear that APACHE_ARROW_VERSION needs to change: its usages seem to be pointing to minimal installs of feast which still refer to 17.0.0

@ntkathole FYI: i've modified lock-python-dependencies-all to pass the args necessary to use the pytorch cpu index

Looks like it is appending +cpu to the requirement explicitly causing issues. I am not able to push to your branch, so created #5662

its not clear that APACHE_ARROW_VERSION needs to change: its usages seem to be pointing to minimal installs of feast which still refer to 17.0.0

I see, right. No need to update since it's still pin to 17.0.0

@ntkathole FYI: this branch also includes a change to the ray / codeflare dependency environment marker. uv lock fails due to its use of python_full_version if this change is not present.