PyTorch Release
This document represents a proposed Change. As part of the Changes process, proposals are publicly announced in order to receive community feedback. This proposal will only be implemented if approved by the Fedora Engineering Steering Committee.
PyTorch Release
This change will bring the first iteration of PyTorch to Fedora.
Owner
This project is owned by the https://fedoraproject.org/wiki/SIGs/PyTorch.
- Name: Tom Rix
- Email: trix@redhat.com
Current status
- Targeted release: Fedora Linux 40
- Last updated: 2024-01-11
- [<will be assigned by the Wrangler> devel thread]
- FESCo issue: <will be assigned by the Wrangler>
- Tracker bug: <will be assigned by the Wrangler>
- Release notes tracker: <will be assigned by the Wrangler>
Detailed Description
The goal of packaging PyTorch for Fedora is to make this open-source machine learning framework easily accessible and seamlessly integrable within the Fedora Linux ecosystem. By providing PyTorch as a packaged software in the Fedora repositories, users gain simplified installation and maintenance processes. This enhances the accessibility of PyTorch for Fedora users, fostering a conducive environment for developers, researchers, and enthusiasts to leverage the capabilities of this powerful machine learning framework. Additionally, packaging PyTorch for Fedora contributes to the broader open-source community by promoting collaborative development and innovation in the field of machine learning on the Fedora platform.
Feedback
The PyTorch SIG meets bi weekly, contributing to the meeting is the best way to engage with what is happening. TBD: Add how to join. The feedback so far has been positive with some high level feature requests. GPU Acceleration More packages that use PyTorch Improving base PyTorch
Benefit to Fedora
This change will introduce PyTorch, a high demand machine learning framework, to Fedora. PyTorch is widely used for tasks such as image and speech recognition, natural language processing, and other artificial intelligence applications, providing a user-friendly interface for building and experimenting with complex machine learning models. This is for CPU (x86_64 and aarch64) only and is the first release for Fedora. The current development effort is focused on AMD GPU acceleration.
Scope
- Proposal owners:
This change does not affect other parts of the distribution and is an isolated change.
- Other developers:
To install use > dnf install python-torch
- Release engineering: N/A (not needed for this Change)
- Policies and guidelines: N/A (not needed for this Change)
- Trademark approval: N/A (not needed for this Change)
- Alignment with Community Initiatives: Yes
Upgrade/compatibility impact
This is the first time this package has been available.
How To Test
The PyTorch code base is very large, start with the public examples described here https://pytorch.org/tutorials/beginner/pytorch_with_examples.html
User Experience
How the user installs PyTorch changes from > pip install torch To > dnf install python-torch
The pip install requires local building which may fail and will not be consistent from machine to machine. Dnf will always succeed and be consistent.
Dependencies
PyTorch is available for Rawhide.
Contingency Plan
- Contingency mechanism: PyTorch is available for Rawhide.
- Contingency deadline: N/A (not a System Wide Change)
- Blocks release? N/A (not a System Wide Change)
Documentation
There is a wealth of online and print documentation available. A good place to start is the main project page https://pytorch.org/
Release Notes
PyTorch 2.1.2 and some supporting packages are available in F40