Improve Cross-Platform Compatibility and Build Process

[sueszli]

Issue

The current build process lacks portability and reproducibility. This makes it harder to collaborate.

By following the build instructions in the CONTRIBUTING.md file, I noticed that the setup.py script is using the same system default compiler for both C and C++ code (e.g., clang on macOS) which breaks the process on MacOS.

After configuring the right compilers per language via the environment variables I noticed that the code isn't portable and breaks on ARM architectures.

Proposed Solution

To enhance portability, reproducibility, and overall project quality, I propose implementing architecture-dependent Docker containers for the build process. This approach will:

1. Ensure consistent build environments across different architectures
2. Simplify the build process for contributors
3. Improve the project's compatibility with a wider range of systems

Benefits

• Increased portability across different architectures, especially ARM
• Enhanced reproducibility of builds
• Simplified onboarding for new contributors
• Expanded reach of the project to more diverse hardware platforms

Related Issues

This pull request addresses the following existing issues:

• Issue #1566
• Issue #4222

Implementation Details

Create Docker containers for each supported architecture.

Testing

I have tested this solution on the following system:

• macOS 14.6.1 23G93 arm64, Mac14,10 (Kernel: 23.6.0) (Apple M2 Pro)

Additional testing on other architectures is recommended before merging.

Acknowledgements

This contribution was supported by ETH-Disco and TUW-DSG.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/8652

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[facebook-github-bot]

Hi @sueszli!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at [email protected]. Thanks!

[sueszli]

CLA has been signed

[NicolasHug]

Hi @sueszli , thanks for the PR.

Sorry, adding docker-based build systems won't be in scope for torchvision.

[sueszli]

I understand - Out of curiosity, may I ask why?

[NicolasHug]

It's because the torchvision builds and release system is based on that of pytorch. Basically it relies on https://github.com/pytorch/test-infra/ and https://github.com/pytorch/builder.

[sueszli]

Thank you! Makes sense – I'll read into it.

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!