-
Notifications
You must be signed in to change notification settings - Fork 967
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
Update torchpippy #2938
Update torchpippy #2938
Conversation
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread. Please note that issues that do not follow the contributing guidelines are likely to be ignored. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for updating @muellerzr ! If you agree with the suggestion below, can you also make the change in our docs ?
src/accelerate/inference.py
Outdated
"`pippy` was not found to be installed on your system. Please " | ||
"install using `pip install torchpippy` or ensure you have at least version 0.2.0" | ||
) | ||
raise ImportError("Using PiPPy requires PyTorch 2.4.0 or later.") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think it's better to refer it as torch.distributed.pipeling
instead of pippy
, so that users are not confused in the future. WDYT ?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes agreed, slight oversight!
Converting to a draft for now as I'm facing a few issues with this. We know that |
323ee76
to
1cb5649
Compare
df1172f
to
1f74cd3
Compare
Update w/ this, all work bar T5 |
Finally merging, torch's integration doesn't support encoder/decoder tracing, so as a result I've raised an error stating to use torchpippy + a lower accelerate version (and as a result, we can't guarantee support) |
This reverts commit 939ce40.
What does this PR do?
Work towards bringing torch 2.4's pippy implementation into Accelerate. Since
torchpippy
itself has not a ton of usage, we're just going to bring it in at once.Fixes # (issue)
Before submitting
Pull Request section?
to it if that's the case.
documentation guidelines, and
here are tips on formatting docstrings.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.