run wheels-publish jobs in Python 3.11 image

[jameslamb]

Contributes to rapidsai/build-planning#7.

Split off from #166.

This proposes upgrading the image used in the wheels-publish workflow to a newer Python version (3.10 -> 3.11).

Benefits of this change

Delays similar updates further into the future (e.g. aws and anaconda-client will drop Python 3.10 support before they drop Python 3.11 support).

How I tested this

This workflow just runs this script: https://github.com/rapidsai/gha-tools/blob/main/tools/rapids-wheels-anaconda.

So just checked that the things that script needs are in the image.

docker run --rm -it rapidsai/ci-wheel:cuda12.0.1-centos7-py3.11 bash

which rapids-wheels-anaconda
# /usr/local/bin/rapids-wheels-anaconda

aws --version
# aws-cli/2.15.7 Python/3.11.6 Linux/5.4.0-148-generic docker/x86_64.centos.7 prompt/off

twine --version
# twine version 4.0.2
# (importlib-metadata: 7.0.1, keyring: 24.3.0, pkginfo: 1.9.6, requests: 2.31.0, requests-toolbelt:
# 1.0.0, urllib3: 2.1.0)

anaconda --version
# anaconda Command line client (version 1.12.2)

I think that's probably sufficient. It'd be easy to revert this if it breaks something... less effort and lower-risk than trying to test on another repo where this wheels-publish only runs on merges to main.

[jameslamb]

This workflow no longer uses twine, as of #154.

Proposing changing this comment to something more future-proof.

[bdice]

I'm okay with this change but I would consider delaying the merge this until we begin work (or complete work) on rapidsai/build-planning#3, currently targeting release 24.04. It doesn't benefit us very much to jump ahead with a new Python version until it's widely used across RAPIDS. Moving all of RAPIDS in lockstep is my preference.

[jameslamb]

Ok makes sense, thanks for that! I agree, this should only be merged as part of rapidsai/build-planning#3, in whatever release that targets.

[bdice]

I converted this to a draft for now. We can reopen once we're moving forward on Python 3.11.

[vyasr]

The Python version used in this jobs can safely be decoupled from everything else, but conceptually I'm fine with keeping things in sync. No rush here.

[jakirkham]

As we are now adding Python 3.11, marking ready for review

[bdice]

I created a test PR for cudf here: rapidsai/cudf#15227

If that passes, we can close it and merge this PR.

[bdice]

I realized that the wheels-publish workflow only runs on build.yaml, not pr.yaml. So rapidsai/cudf#15227 isn't a sufficient test. There isn't a good way to test this because it only triggers when PRs are merged (or during nightly tests). I'm going to go ahead and merge it so we can get feedback from nightlies and revert tomorrow if necessary.