Skip to content
This repository has been archived by the owner on Sep 25, 2023. It is now read-only.

Dependency differences between Conda & wheel #571

Open
jakirkham opened this issue May 9, 2023 · 3 comments
Open

Dependency differences between Conda & wheel #571

jakirkham opened this issue May 9, 2023 · 3 comments

Comments

@jakirkham
Copy link
Member

In the Conda recipe, these dependencies are listed

run:
- cupy >=12.0.0
- numba >=0.49.0
- numpy >=1.19,<1.24
- python
- scipy >=1.6.0

However the wheel only lists one of those

INSTALL_REQUIRES = ["numba"]

Should these be synced to have the same lists?

@jakirkham
Copy link
Member Author

cc @vyasr (for vis)

@awthomp
Copy link
Member

awthomp commented May 9, 2023

The Python wheel should include the dependencies listed in the conda meta.yaml recipe. cupy, numpy, scipy, and numba are all hard dependencies for cuSignal.

@vyasr
Copy link
Contributor

vyasr commented May 9, 2023

Ideally they should be synced, but in practice until we start promising that our packages are pip installable it doesn't really matter. Pretty much all of RAPIDS had the wrong dependencies listed in setup.py/setup.cfg/pyproject.toml until we actually started making wheels. It would be nice to fix, but it's hard to synchronize without some process to actually build and test using those dependencies (unless you update to producing pyproject.toml dependencies using rapids-dependency-file-generator, which would also require transitioning this package to being pyproject.toml based).

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants