Skip to content

fix(build): Handle namespace packages for PyPI CUDA detection #2580

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.

Sign up for GitHub

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

Merged
ksivaman merged 2 commits into from
Feb 12, 2026
Merged

fix(build): Handle namespace packages for PyPI CUDA detection #2580

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d8b7216
fix: handle nvidia namespace packages where __file__ is None
sbhavani Feb 9, 2026
a18edc8
Merge branch 'main' into fix/nccl-pypi-detection
ksivaman Feb 11, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 6 additions & 4 deletions transformer_engine/common/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -245,11 +245,13 @@ def _nvidia_cudart_include_dir() -> str:
return ""

# Installing some nvidia-* packages, like nvshmem, create nvidia name, so "import nvidia"
# above doesn't through. However, they don't set "__file__" attribute.
if nvidia.__file__ is None:
return ""
# above doesn't throw. However, they don't set "__file__" attribute.
if nvidia.__file__ is not None:
nvidia_root = Path(nvidia.__file__).parent
else:
nvidia_root = Path(nvidia.__path__[0]) # namespace package
Comment on lines +250 to +252
Copy link
Contributor

@greptile-apps greptile-apps bot Feb 11, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Unconditional __path__[0] indexing
If nvidia.__path__ exists but is empty (or contains a non-path-like entry), nvidia.__path__[0] will raise IndexError/TypeError, which would regress the “clear error message” intent. Consider guarding for emptiness and using the same explicit RuntimeError path.

Suggested change
nvidia_root = Path(nvidia.__file__).parent
else:
nvidia_root = Path(nvidia.__path__[0]) # namespace package
nvidia_dir = list(nvidia.__path__)[0]

→ check nvidia.__path__ is truthy/non-empty before indexing and raise the intended error when it isn’t.


include_dir = Path(nvidia.__file__).parent / "cuda_runtime"
include_dir = nvidia_root / "cuda_runtime"
return str(include_dir) if include_dir.exists() else ""


Expand Down
Loading