-
Notifications
You must be signed in to change notification settings - Fork 0
What is considered a SEV?
Why do we file SEVs?
- Alert devs that something is currently wrong with the CI and they may see failures that they’re not responsible for
- Track developer facing issues so that we can appropriately prioritize fixing (without SEVs these can get forgotten)
When there is a widespread CI breakage that is not resolvable by doing a revert (e.g. it’s an issue with the underlying CI infra), you should file a SEV using the CI SEV issue template.
- If it’s bad enough that people multiple people will ask us about it on any of our contributor communication channels
- If it affects more than ~10% of users and might last over 30 minutes, it’s a SEV
- If we’re expecting people to force push to work around it, it’s a SEV
If the above still leaves it ambiguous, we default to considering it a SEV.
If the sev is severe enough (generally defined as "the CI is likely to miss legitimate failures", the template includes instructions on how to make it merge blocking to prevent contributors from accidentally merging broken code.
HUD and the Dr. CI comment at the top of the PR will both explain that there's a SEV ongoing and link to the active sev.
If the SEV is marked as merge blocking, @pytorchbot will default to preventing PRs from being merged
PyTorch presented to you with love by the PyTorch Team of contributors
- Install Prerequisites and Dependencies
- Fork, clone, and checkout the PyTorch source
- Build PyTorch from source
- Tips for developing PyTorch
- PyTorch Workflow Git cheatsheet
- Overview of the Pull Request Lifecycle
- Finding Or Creating 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)