Skip to content
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

Update setup.py: Make more requirements optional #4133

Merged
merged 7 commits into from
Feb 25, 2022

Conversation

adelizer
Copy link
Contributor

Moving the following packages to full_install_requires:

  • yacs
  • hydra-core
  • class-resolver>=0.3.2
  • googledrivedownloader

Keeping only essential packages in install_requires will enable torch-geometric to be installed in secure environments with strict package management constraints.

@adelizer
Copy link
Contributor Author

As the community for torch-geometric grows, this PR will make it more accessible to be installed and used within highly secure environments. This is the case for example at Adyen https://www.adyen.com/ where we would like to start using torch-geometric but couldn't install some of the dependencies due to security concerns.

@rusty1s
Copy link
Member

rusty1s commented Feb 24, 2022

Thanks for the PR. I'm happy to move some of these dependencies to optional ones. We would need to add a new CI test to ensure that PyG can still successfully run even when dropping these dependencies. Furthermore, I wonder what are the security concerns of these packages? These are all pretty popular packages.

@adelizer
Copy link
Contributor Author

Sure, I know of this tool https://snyk.io/advisor/python being used to check the health of different packages and in the case of googledrivedownloader it is definitely low as well as for yacs. I can modify the PR to only move these 2 if you suggest so.
Furthermore, I can also help with the new CI test if needed.

@rusty1s
Copy link
Member

rusty1s commented Feb 25, 2022

@adelizer I added a GitHub action to test a minimal installation (currently checks if imports work as expected). As the test fails, we need to make the dependencies truly optional before we can merge this PR.

@rusty1s rusty1s changed the title Update setup.py Update setup.py: Make more requirements optional Feb 25, 2022
@rusty1s
Copy link
Member

rusty1s commented Feb 25, 2022

All issues are now resolved. Thank you for the PR!

@rusty1s rusty1s merged commit cd04dd2 into pyg-team:master Feb 25, 2022
@adelizer
Copy link
Contributor Author

@rusty1s Thanks a lot!! Any proposed date for the next release with this change included?

@rusty1s
Copy link
Member

rusty1s commented Feb 25, 2022

Next release is scheduled around mid March. Let me know if this is early enough for you.

@adelizer
Copy link
Contributor Author

If there is an option for an earlier one that would be greatly helpful!

@rusty1s
Copy link
Member

rusty1s commented Feb 25, 2022

Do you need to install from PyPi or does installing via GitHub work as well?

pip install git+https://github.com/pyg-team/pytorch_geometric.git

AFAIK, PyTorch 1.11 is coming soon, and we would like to wait for it before releasing any new version.

@adelizer
Copy link
Contributor Author

Great, thanks for explaining! I will wait for mid March then since it is not possible to install directly from GitHub. As in our use case all packages are mirrored on an internal PyPI

@rusty1s
Copy link
Member

rusty1s commented Mar 1, 2022

Hey @adelizer, can you reach out to me on matthias.fey@tu-dortmund.de? I would like to get in touch with you about the use-cases of PyG @ Adyen.

@rusty1s
Copy link
Member

rusty1s commented Mar 12, 2022

@adelizer Just released PyG 2.0.4, see here :)

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

Successfully merging this pull request may close these issues.

None yet

2 participants