-
Notifications
You must be signed in to change notification settings - Fork 1.2k
feat: add ignored_params support for fsdp2 #3731
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
Conversation
Signed-off-by: Mehant Kammakomati <mehant.kammakomati2@ibm.com>
Signed-off-by: Mehant Kammakomati <mehant.kammakomati2@ibm.com>
Signed-off-by: Mehant Kammakomati <mehant.kammakomati2@ibm.com>
Signed-off-by: Mehant Kammakomati <mehant.kammakomati2@ibm.com>
Hello, thank you for the PR. Can you also include it in the tests? |
Hi 👋🏻 @S1ro1 Yes I will update the PR with the test. |
Signed-off-by: Mehant Kammakomati <mehant.kammakomati2@ibm.com>
@S1ro1 Have added it. |
@S1ro1 Should I bump up the torch version |
@S1ro1 appreciate any update ? |
@kmehant Sorry, missed the update. I think it should be handled separately, something in post_init of fsdp plugin, along the lines of |
Signed-off-by: Mehant Kammakomati <mehant.kammakomati2@ibm.com>
@S1ro1 thank you for the response, does the current change sound good? Also, do you think we should expose ignored_params as an option for the user to provide apart from having ignored_modules as a proxy? |
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.
LGTM
@kmehant I think let's see if this gets some usage, can you think of some common use-cases where you'd like to use ignored_params directly? I feel like |
Signed-off-by: Mehant Kammakomati <mehant.kammakomati2@ibm.com>
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. |
I don't have a usecase for ignored_params we can keep it for future.
We have a specific use case for this, this PR will be helpful. |
Failed test seems unrelated |
What does this PR do?
This allows reusing
ignored_modules
user facing FSDP API for FSDP v2 to setignored_params
.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.
@SunMarc @zach-huggingface @S1ro1