From Fedora Project Wiki
mNo edit summary
Tag: Manual revert
No edit summary
 
Line 36: Line 36:
* [https://gitlab.com/fedora/sigs/ai-ml/ Fedora GitLab AI/ML] - Git repos for packages, issue tracking, external contributions, automation.
* [https://gitlab.com/fedora/sigs/ai-ml/ Fedora GitLab AI/ML] - Git repos for packages, issue tracking, external contributions, automation.
* [[SIGs/PyTorch/packagingStatus|PyTorch Packaging Status]] - List of package dependencies, current packaging status, and coordination on who is working on which package.
* [[SIGs/PyTorch/packagingStatus|PyTorch Packaging Status]] - List of package dependencies, current packaging status, and coordination on who is working on which package.
* [https://github.com/pytorch/pytorch/actions/workflows/generated-linux-binary-manywheel-nightly.yml Upstream Wheel Builds] - Reference builds from upstream.


=== Buildsys ===
=== Buildsys ===

Latest revision as of 18:00, 20 July 2024

Help wanted!
You can help with this stuff and help define the future of PyTorch in Fedora! Join us on #fedora-ai-ml[?] on Libera IRC and #ai-ml:fedoraproject.org on Matrix

PyTorch Special Interest Group (SIG)

PyTorch is a machine learning framework based on the Torch library, used for applications such as computer vision and natural language processing, originally developed by Meta AI and now part of the Linux Foundation umbrella. It is free and open-source software released under the modified BSD license.

The intent of the PyTorch SIG is to bring together all people that are interested in AI/ML/HC/HPC in Fedora, be it packagers, developers and/or users and working on improving the overall experience. We intend on pushing forward native PyTorch support in Fedora. This includes collaborating with other groups to enable hardware acceleration and needed package dependencies.

Mission

  • To package and maintain all required dependencies for a seamless PyTorch experience in Fedora.
  • To make debugging issues reported against accelerator components easier to reproduce.
  • To coordinate hardware enablement efforts that will benefit all of Fedora.

Licensing

  • Pretrained models or datasets will be treated as content.

Get Started

TBD - instructions on how to quickly fire up PyTorch and do something interesting; this should likely be a hosted notebook somewhere (maybe colab or some other externally willing provider) for demo, and then something like https://pytorch.org/get-started/locally/ until we get everything working packaged; consider publishing a container based on fedora with everything possible provided by packages and then the remaining steps automated (hw acceleration might be challenging, but is possible)

Get Involved

TBD - instructions on how to get involved

  • Join the AI/ML SIG FAS Group by introducing yourself on Matrix and sharing what packages you would like to focus on. Please note you will need to be an existing packager for group membership.

Communication

Let's get together and make this happen. Join the discussion:

Meetings

The meeting schedule, purpose, technology, and attendees is still being developed. The latest SIG meeting will be listed on https://calendar.fedoraproject.org/SIGs/.

Resources

Buildsys

We will use COPR to coordinate work and compute resources to start.

  • Main SIG COPRs/Repos:
  1. AI/ML SIG GitLab Repository
  2. AI/ML SIG COPR Group
  • Contributor/Community COPRs/Repos:
  1. https://copr.fedorainfracloud.org/coprs/rezso/ML/
  2. https://copr.fedorainfracloud.org/coprs/g/neurofedora/neurofedora-extra/
  3. https://copr.fedorainfracloud.org/coprs/jmontleon/pytorch/builds/
  4. https://github.com/trixirt/pytorch-fedora -- A bootstrapping specfile from Tom Rix