Let nvidia-ml-py and pynvml 12.* coexist

[jakirkham]

Beginning with pynvml version 12.0.0, the pynvml library and package now depend on nvidia-ml-py for core functionality and merely provide a few extras on top of nvidia-ml-py. Given this, these two packages no longer clobber each other and instead nvidia-ml-py needs to be coinstallable with pynvml. Changes were made to both feedstocks to support this (as referenced below). This repodata patch extends this to nvidia-ml-py version 12.*.

References:

• Remove pynvml module and convert to a "meta package" gpuopenanalytics/pynvml#57
• pynvml v12.0.0 pynvml-feedstock#18
• Allow pynvml and nvidia-ml-py to be coinstalled for 12.* versions nvidia-ml-py-feedstock#24

---

Checklist

• Used a static YAML file for the patch if possible (instructions).
• Only wrote code directly into generate_patch_json.py if absolutely necessary.
• Ran pre-commit run -a and ensured all files pass the linting checks.
• Ran python show_diff.py and posted the output as part of the PR.
• Modifications won't affect packages built in the future.

[jakirkham]

cc @conda-forge/nvidia-ml-py @conda-forge/pynvml (for awareness)

[jakirkham]

Here is the repodata patch diff from CI:

noarch::nvidia-ml-py-12.555.43-pyhd8ed1ab_0.conda
noarch::nvidia-ml-py-12.535.77-pyhd8ed1ab_0.conda
noarch::nvidia-ml-py-12.535.108-pyhd8ed1ab_0.conda
noarch::nvidia-ml-py-12.560.30-pyhd8ed1ab_0.conda
noarch::nvidia-ml-py-12.550.52-pyhd8ed1ab_0.conda
noarch::nvidia-ml-py-12.535.133-pyhd8ed1ab_0.conda
-    "pynvml ==9999999999"
+    "pynvml ~=12.0"

[hmaarrfk]

what does ~=12.0 mean?

[jakirkham]

This is equivalent to pynvml >=12.0,<13.0a

More details in the package match specification docs

xref to similar discussion: conda-forge/pynvml-feedstock#18 (comment)

[jakirkham]

@conda-forge-admin , please relint

(looks like the linter errored during installation)

[conda-forge-admin]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe/meta.yaml) and found it was in an excellent condition.

[hmaarrfk]

LGTM

[jakirkham]

Thanks Mark and Keith! 🙏

[jaimergp]

I remember we were patching =~ out because it would break old conda versions. Also mamba v2 seems to have some trouble with it. Might be safer to use the equivalent >=x.y,x.*.

[jakirkham]

How old of a version?