Introduction to TensorFlow SIG Build
TensorFlow SIG Build is a community-driven initiative that focuses on enhancing the build process of TensorFlow, a popular open-source machine learning framework. This comprehensive project serves as a repository for resources, guides, and tools that are contributed by and for the community to streamline TensorFlow build processes.
Group Overview
Contributing to the Project
TensorFlow SIG Build thrives on the contributions from its community. Being an open-source project, it relies heavily on public involvement in the form of contributions, bug-fixes, and documentation enhancements. Participants are encouraged to review the contribution guidelines to understand how to effectively contribute to the project. The project emphasizes adhering to TensorFlow's code of conduct to maintain a welcoming and productive community environment.
Community Interaction
- Mailing List: A public mailing list is available for discussions. This platform allows community members to stay updated and participate in conversations relevant to TensorFlow SIG Build.
- Monthly Meetings: The community hosts regular monthly meetings, and the notes from these meetings are accessible to all. Interested contributors can join the mailing list to receive invitations to these meetings.
Licensing
TensorFlow SIG Build is distributed under the Apache License 2.0, permitting broad use and modification with proper attribution.
Project Showcase
The project showcases various resources and artifacts aimed at simplifying the TensorFlow build process. The community can easily add their projects by following the guidelines available.
Docker Resources
- TF SIG Build Dockerfiles: These are standard Dockerfiles used for building TensorFlow, both internally by Google and by the community.
- TensorFlow Runtime Dockerfiles: These offer simple solutions for running TensorFlow, including variants for Jupyter.
- Manylinux 2014 Docker Images: A build environment for TensorFlow packages.
- Distroless Dockerfiles: Provide smaller, more efficient TensorFlow images compared to official releases.
- DevInfra Windows RBE: A static snapshot of TensorFlow's Windows Remote Build Execution images.
Language Bindings
- Golang Install Guide: Offers clear documentation for installing Go bindings for TensorFlow.
Platform Builds
- ppc64le Builds: Resources for building TensorFlow on the ppc64le architecture.
- Raspberry Pi Builds: Historic documentation for building TensorFlow on Raspberry Pi, seeking an owner for maintenance.
- WSL2 GPU Guide: A helpful guide to enable GPU support for TensorFlow running on WSL2 virtual machines.
Other Work-in-Progress
- Directory Template: An example of how to structure project descriptions.
- TF OSS Dashboard: Provides continuous status updates on TensorFlow GitHub commits.
- Tekton CI: An experimental directory exploring the use of Tekton CI with TensorFlow.
Community Supported TensorFlow Builds
TensorFlow SIG Build highlights several community-supported builds, which expand TensorFlow's reach to platforms not officially supported by the TensorFlow team. These builds are invaluable for testing, packaging, and running TensorFlow on diverse hardware.
TensorFlow Builds by Platform
Notable community builds include:
- AMD: Provides nightly and stable builds for Linux AMD ROCm GPUs and ZenDNN CPU plugins.
- IBM: Offers builds for Linux on ppc64le, s390x CPUs, and NVIDIA GPUs across various versions.
- Intel: Supplies builds optimized for Intel CPUs with oneDNN support.
- Linaro: Focuses on supporting ARM architectures, with stable releases for Linux aarch64 CPUs.
TensorFlow Containers
Several Docker container images are maintained by various contributors for efficient TensorFlow deployment:
- Linaro: Images optimized for ARM Neoverse-N1 CPUs.
- AMD: Provides container images tailored for Linux ROCm GPU environments.
- Intel: Offers optimized containers for Intel CPUs.
Each of these community contributions significantly enhances the flexibility and accessibility of TensorFlow across many computing environments. This collaborative effort underscores the vital role that the community plays in expanding TensorFlow's capabilities beyond its official offerings.