Fix np.dtype overload signature drift#797
Conversation
Match the CUDA np.dtype overload parameters to NumPy's public signature so signature-compatibility checks pass on environments where dtype exposes align/copy/kwargs parameters. Co-authored-by: Cursor <cursoragent@cursor.com>
|
Automatic reviews are disabled for this repository. |
|
Auto-sync is disabled for ready for review pull requests in this repository. Workflows must be run manually. Contributors can view more details about this message here. |
|
/ok to test |
Keep the overload argument names aligned with NumPy while avoiding VAR_KEYWORD in the implementation signature, which breaks CUDA typing on calls like np.dtype("O").
Co-authored-by: Cursor <cursoragent@cursor.com>
|
/ok to test |
Rename CUDA overload parameters for np.dot and np.vdot to match NumPy's public argument names so signature-compatibility checks pass across NumPy variants. Co-authored-by: Cursor <cursoragent@cursor.com>
|
/ok to test |
brandon-b-miller
left a comment
brandon-b-miller
left a comment
There was a problem hiding this comment.
cuDF failures are just a broken nightly I think, lets rerun CI in the morning
Sorry @cpcloud this was wrong, I forgot we're on the stable release of rapids at this point and it looks like cupy 14 is breaking that release. Can we try installing cupy<0.14.0 in the test script before the tests are run? https://github.com/NVIDIA/numba-cuda/blob/main/ci/test_thirdparty_cudf.sh We can undo this later after the cupy patch version comes out. |
|
/ok to test |
2 participants
Summary
np.dtypeoverload parameters with NumPy (dtype,align,copy) while avoiding unsupported**kwargsin typing templatesnp.dotandnp.vdotoverload parameter names with NumPy (a,b,out) so signature-compatibility checks pass across NumPy variantsExtracted from #795 to keep the issue-718 DCE fix focused.