Here are the remaining items we should do before the PyTorch 1.13 release. Note that PyTorch 1.13 has cut a release branch, so if we are going to find bugs, we should find and fix them soon. The absolute deadline is 10/21.
P0
[x] AOTAutograd tutorial is broken. It doesn't run and appears to give silently incorrect results. @Chillee
[x] TorchBench microbenchmark regression suite, test new release for regressions @zou3519
[x] functorch 1.13 docs (this is separate from pytorch docs...) @zou3519
[x] PyTorch release notes and release blurb @zou3519
[x] check version numbers @zou3519
[ ] Test functorch 1.13.0 pip binary (pip install functorch, see if torch of correct version gets installed) @zou3519
[ ] Publish functorch 1.13.0 binary
[ ] Publish functorch 1.13.0 release notes
P1 (Nice to have, especially if they catch critical bugs):
[ ] Test functorch over torchbench models @kshitij12345 Tests do not need to go into the release; we should try to fix any hi-pri bugs we find as a result of this.
Here are the remaining items we should do before the PyTorch 1.13 release. Note that PyTorch 1.13 has cut a release branch, so if we are going to find bugs, we should find and fix them soon. The absolute deadline is 10/21.
P0
P1 (Nice to have, especially if they catch critical bugs):