Finding Or Reporting Issues
The majority of open source contributions come from people scratching their own itches. However, if you don't know what you want to work on, or are just looking to get more acquainted with the project, look through the Issue Tracker. The Issue Tracker lists open (and closed) bug reports or requests for new features. Issues that are confirmed by other contributors tend to be better to investigate. Every triaged issue contains [PyTorch AutoLabel Bot | labels] that provide more information about the reported problem.
-
The "good first issue" label indicates issues that should feel approachable to fix, and typically don't need wide-ranging changes in the codebase (feel free to leave a comment if you think the classification no longer applies).
-
The "actionable" label denotes issues that have a solution identified but not yet implemented.
If you decide to help out, please leave a comment in the issue thread to start a conversation with us.
Before filing a new issue, check if the problem has already been reported on the tracker. This will help you come up to speed on the latest state of the issue, and you might find a solution if it has been resolved!
If you don't see your bug/feature already reported, please open a new issue. Choose a template that closely matches your proposal. Each template contains instructions and questions that make it easier for maintainers to review and triage.
You can tag the module maintainers if you think the issue should be brought to their attention. Use the CODEOWNERS and the Maintainers List to know who should be tagged for specific areas.
Once an issue is raised, someone from the PyTorch Triaging team will take a look, check for validity, and assign appropriate labels. PyTorch maintainers review new issues almost daily.
No, but please leave a note on the issue when working on it so others are aware.
We don't "lock" issues and only allow a single person to work on them. Even if an issue is assigned you are free to work on it and submit a pull request for it. If there are multiple pull requests addressing the same issue then the one that is the highest quality (or that came first, if multiple have the same quality) will be accepted.
- Install Prerequisites
- Fork, clone, and checkout the PyTorch source
- Install Dependencies
- Build PyTorch from source
- Tips for developing PyTorch
- PyTorch Workflow Git cheatsheet
- Overview of the Pull Request Lifecycle
- Finding Or Reporting Issues
- Pre Commit Checks
- Create a Pull Request
- Typical Pull Request Workflow
- Pull Request FAQs
- Getting Help
- Codebase structure
- Tensors, Operators, and Testing
- Autograd
- Dispatcher, Structured Kernels, and Codegen
- torch.nn
- CUDA basics
- Data (Optional)
- function transforms (Optional)