DOC: Move CUDA user guide from cuda-feedstock to conda-forge docs #2702
Conversation
✅ Deploy Preview for conda-forge-previews ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
docs/maintainer/knowledge_base.md
Outdated
| not require e.g. extra math libraries like cuBLAS or cuSPARSE). | ||
|
|
||
|
|
||
| #### Compatability between the `conda-forge` vs `nvidia` channels |
There was a problem hiding this comment.
This section will soon become a historic detail that maybe is not too relevant in the future. 12.5 is almost two years old. How do you feel about putting this under a <details> collapsible panel?
There was a problem hiding this comment.
I agree that this section is very long and mostly irrelevant today. I will try to figure out how to mark the section as collapsable.
Co-authored-by: h-vetinari <h.vetinari@gmx.com>
|
Thanks for feedback, @h-vetinari and @kkraus14 |
jakirkham
left a comment
jakirkham
left a comment
There was a problem hiding this comment.
Thanks Daniel! 🙏
Also thank you for everyone helping review
Had a few notes on the history, which provide a bit more context
Otherwise this looks pretty good
| packages. Due to the scale of the reorganization, the CUDA 12.0, 12.1, and 12.2 releases | ||
| also involved numerous additional fixes to the packaging structure to better integrate them | ||
| in the conda ecosystem. Due to the number of such changes that were required and the focus | ||
| on improving the quality of these installations, during this time period no corresponding |
There was a problem hiding this comment.
| on improving the quality of these installations, during this time period no corresponding | |
| on improving the quality of these installations during this time period, no corresponding |
There was a problem hiding this comment.
Disagree. "During this time period" is describing the lack of updates on the nvidia channel. It's not describing focus on improving quality.
There was a problem hiding this comment.
In that case think we should try to simplify this sentence. Possibly by breaking it into two sentences or shortening what is in it
|
|
||
| Going forward, CUDA packages on the `conda-forge` and `nvidia` channels should be expected | ||
| to remain compatible. | ||
|
|
There was a problem hiding this comment.
Think it is worth noting that this is becoming a historical footnote mostly relevant for older packages
| As of December 2025, conda-forge builds packages supporting CUDA 12.6+. So freshly built packages in conda-forge should be compatible with both `nvidia`, `conda-forge`, and `defaults` channels. | |
There was a problem hiding this comment.
conda-forge doesn't have any compatibility guarantees with other channels, so I'm hesitant to imply otherwise.
There was a problem hiding this comment.
Mostly wanted to note that Anaconda has completed other equivalent work with defaults
Am flexible on how we do that
What would you propose?
There was a problem hiding this comment.
The most we can say is the the CTK libraries are the same across the channels because we (NVIDIA) only allow binary redists of these packages. We cannot state that these three channels are compatible because it's not true; there exist non-CTK packages on these channels which break compatibility.
Co-authored-by: jakirkham <jakirkham@gmail.com>
5 participants
PR Checklist:
docs/orcommunity/, you have added it to the sidebar in the corresponding_sidebar.jsonfileMoves the end-users guides (not the maintainers guide) from cuda-feedstock to these docs.
Addresses part of #2634