start using CUDA 12.5, Python 3.11 images in nightly docs#398
start using CUDA 12.5, Python 3.11 images in nightly docs#398jacobtomlinson merged 3 commits intorapidsai:mainfrom jameslamb:nightly-images
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
|
Thanks James! 🙏 Seems like a good idea 👍 Let's see what @jacobtomlinson thinks 🙂 |
jacobtomlinson
left a comment
There was a problem hiding this comment.
This seems like a good change.
I just want to double check my understanding about backward compatibility. If a platform gives us nodes with CUDA 12.0, are we fine to use the CUDA 12.5 images? My understanding is that there is backward compatibility within major versions these days, but I just want to check that is true.
|
Yes that is still how things work As long as they supply hardware that supports CUDA 12 and there is a compatible driver version (either provided by them or us via a container), then it should work Please see this page for driver versions |
|
Thanks @jakirkham that's really helpful! |
|
Thanks Jacob and James! 🙏 |
Contributes to #377.
Proposes using CUDA 12.5, Python 3.11 container images for the nightly docs. This should make it easier to test with these versions during release testing for 24.08. And hopefully that testing goes well and we can move on to setting the stable versions to these images as well.
CUDA 12.5
rapidsai/baseandrapidsai/notebooksimages were added in rapidsai/docker#689, as part of rapidsai/build-planning#73.