Skip to content
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

MS-AMP support (w/o FSDP) #3093

Merged
merged 6 commits into from
Sep 10, 2024
Merged

MS-AMP support (w/o FSDP) #3093

merged 6 commits into from
Sep 10, 2024

Conversation

muellerzr
Copy link
Collaborator

@muellerzr muellerzr commented Sep 9, 2024

What does this PR do?

Fixes up the MS-AMP integration to include:

  • DeepSpeed support (1-2)
  • Integration suite
  • misaligned init

FSDP is unsupported while Azure/MS-AMP#181 is in limbo.

Closes #2143
Closes #2971
Closes #3050

Before submitting

  • This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
  • Did you read the contributor guideline,
    Pull Request section?
  • Was this discussed/approved via a Github issue or the forum? Please add a link
    to it if that's the case.
  • Did you make sure to update the documentation with your changes? Here are the
    documentation guidelines, and
    here are tips on formatting docstrings.
  • Did you write any new necessary tests?

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.

@BenjaminBossan @SunMarc

@HuggingFaceDocBuilderDev

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.

Copy link
Member

@BenjaminBossan BenjaminBossan left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for addressing these issues. Overall, this looks good, I have a couple of comments but no big ones.

Nice benchmarks too, I didn't check all the details, as they're based on the existing ones AFAICT.

src/accelerate/utils/dataclasses.py Show resolved Hide resolved
src/accelerate/accelerator.py Outdated Show resolved Hide resolved
src/accelerate/accelerator.py Outdated Show resolved Hide resolved
src/accelerate/accelerator.py Show resolved Hide resolved
src/accelerate/accelerator.py Outdated Show resolved Hide resolved
@muellerzr muellerzr force-pushed the muellerzr-msamp-fixings branch from 5289332 to 519c417 Compare September 10, 2024 14:51
Copy link
Member

@BenjaminBossan BenjaminBossan left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the fixes, LGTM. As mentioned internally, I could not test it though.

Copy link
Member

@SunMarc SunMarc left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for adding the support with Deepspeed ! Nice benchmarks for fp8. Just a small nit

src/accelerate/accelerator.py Outdated Show resolved Hide resolved
@muellerzr muellerzr merged commit d5b7b70 into main Sep 10, 2024
28 checks passed
@muellerzr muellerzr deleted the muellerzr-msamp-fixings branch September 10, 2024 16:25
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Wrong Initialization of msamp in accelerate MS-AMP support for DeepSpeed Feature Request: Support MS-AMP
4 participants