-
Notifications
You must be signed in to change notification settings - Fork 204
Switch to cuda-toolkit metapackage for wheel dependencies with streamlined documentation #883
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
Co-authored-by: leofang <[email protected]>
cuda-toolkit
metapackage to constraint Python wheel dependenciesThere was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
- cuda_bindings: merge nvcc,nvrtc,nvjitlink,nvvm into single line - cuda_pathfinder: merge all extras except cufile (platform conditional) for cu12/cu13 Co-authored-by: leofang <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Co-authored-by: leofang <[email protected]>
/ok to test 6a12072 |
This comment has been minimized.
This comment has been minimized.
…pathfinder Co-authored-by: leofang <[email protected]>
/ok to test 5d4e179 |
…tes grammar Co-authored-by: leofang <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I added the two "the" manually: commit cd7c41c
Thanks, guys! Since the CI was green and last two commits are doc change-only, let me admin-merge. |
|
This PR migrates all NVIDIA wheel dependencies from individual packages to the new
cuda-toolkit
metapackage, providing safer version constraints, and adds streamlined documentation for the changes.Changes Made
Dependency Migration:
cuda_bindings/pyproject.toml
: Updated optional dependencies in the[all]
section to usecuda-toolkit[nvrtc,nvjitlink,nvvm]==13.*
(no longer includesnvcc
as it was only needed for libNVVM access, which now has its own dedicated wheel)cuda_core/pyproject.toml
: Updated test dependencies to usecuda-toolkit[cudart]
with appropriate version constraints and addedtest-cu13
dependency now thatcupy-cuda13x
is availablecuda_pathfinder/pyproject.toml
: Migrated from individual wheels to equivalentcuda-toolkit
extras, includingnvcc
for both cu12 and cu13 sectionsDocumentation Updates:
cuda-toolkit
metapackagecuda-version
metapackage to control CUDA Toolkit componentsImportant Notes
NVCC Compiler Changes:
pip install cuda-python[all]
. Previously, NVCC was included only to access the NVVM library, but CUDA 13 now provides NVVM as a separate wheel (nvidia-nvvm
). Users who need the NVCC compiler should explicitly install it:pip install nvidia-cuda-nvcc
Benefits
Fixes #853.
✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.