New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[v.1.10.0] Release Tracker #65438
Comments
Link to landed master PR: N/A @zhouzhuojie: merged |
AveragedModel's update_parameter() doesn't update the parameters correctly in certain cases. This PR fixes that. Look at #65495 (comment) and #65495 (review) for more details. |
a likely regression - not triaged - #66086 |
Hi, we would like to get Intel Extension to Pytorch tutorial into 1.10 release. Could you please merge it in? @vitaly-fedyunin and @gottbrath give me this link. Thanks. @malfet |
This issue is already have 1.10 milestone |
@Jianhui-Li can you please share link to the PR? Has it been merged into the main branch yet? |
Sorry to forget the link pytorch/tutorials#1702 |
Sorry, but #66712 might be a blocking issue for the release. I had introduced a bug that I fixed in #66703. 😞 |
Hmm, this is indeed unfortunate, but what's the performance impact? Are AVX2 or default kernels will be selected in that case? |
AVX2 |
In that case it does not sound like a release blocker: i.e. users would not benefit from AVX512 perf boost, but CPU performance would be no different (and sometimes faster), than in PyTorch-1.9.0 |
Thanks, @malfet! Yes, the primary issue is that without fixing the aforementioned bug, many users might be misled about the ATen CPU capability. BTW, please confirm if it's possible to release another binary that also has AVX512 kernels. |
hi @malfet : about the AVX-512 support, since it is a major feature added after 1.9 by @imaginary-person , will you highlight it for the 1.10 release? If so, having it enabled by default in the 1.10 binary release sounds more reasonable? |
Did you guys just accidentally release 1.10? It can be installed through conda:
However, neither the releases page nor the 1.10.0 milestone nor the 1.10 release tracker nor pypi indicate the update shipped yet. |
Closing this issue since 1.10.0 has been released! Thanks everyone in this thread for all of the contributions that made 1.10.0 happen! |
Started another torchaudio build to account for missing cuda-113 windows binaries: https://app.circleci.com/pipelines/github/pytorch/audio/8064/workflows/3c737d4b-72fa-4bb1-b5c0-90fa0c82c21a |
Started torchvision build to fix packaging problem for cuda-113 linux wheels: |
We cut a release branch for the 1.10.0 release.
Our plan from this point from this point is roughly:
This issue is for tracking cherry-picks to the release branch.
Cherry-Pick Criteria
Phase 1 (until 10/7/21):
Only low-risk changes may be cherry-picked from master:
Any other change requires special dispensation from the release managers (currently @malfet, @seemethere, @gchanan, @zhouzhuojie ). If this applies to your change please write "Special Dispensation" in the "Criteria Category:" template below and explain.
Phase 2 (after 10/7/21):
Note that changes here require us to rebuild a Release Candidate and restart extended testing (likely delaying the release). Therefore, the only accepted changes are Release-blocking critical fixes for: silent correctness, backwards compatibility, crashes, deadlocks, (large) memory leaks
Changes will likely require a discussion with the larger release team over VC or Slack.
Cherry-Pick Process
Ensure your PR has landed in master. This does not apply for release-branch specific changes (see Phase 1 criteria).
Create (but do not land) a PR against the release branch.
Make a request below with the following format:
NOTE: Our normal tools (ghstack / ghimport, etc.) do not work on the release branch.
See HUD 1.10 or #65434 for release branch CI status
The text was updated successfully, but these errors were encountered: