Add Python 3.15 to the test matrix#2108
Conversation
| - { ARCH: 'amd64', PY_VER: '3.15', CUDA_VER: '13.2.1', LOCAL_CTK: '1', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest' } | ||
| - { ARCH: 'amd64', PY_VER: '3.15t', CUDA_VER: '13.2.1', LOCAL_CTK: '1', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest' } |
There was a problem hiding this comment.
It would be nice to not add additional jobs if possible where can we take existing jobs in the matrix and update their Python version without introducing coverage concerns?
|
Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually. Contributors can view more details about this message here. |
|
/ok to test |
|
/ok to test |
|
/ok to test |
|
It seems we need to build PY315 in the https://github.com/NVIDIA/cuda-python/actions/runs/26123422715/job/76831314547?pr=2108#step:38:76 |
Backport of #2108 to the 12.9.x branch so that cuda-bindings 12.9.x wheels are available for Python 3.15, which the main branch needs in order to build cuda-core. - Add 3.15 / 3.15t to the build matrix - Upgrade cibuildwheel to v4.0.0rc1 (required for cpython-prerelease) - Enable cpython-prerelease in CIBW_ENABLE - Add amd64-only test entries for 3.15 / 3.15t (CUDA 12.9.1)
This is done in #2133. Need to merge it first and then apply the learnings there in this PR. |
Add Python 3.15 to the test matrix (Linux x86 only for now) so that we are ready for any issues by October.