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.

Sign up for GitHub

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

Jump to bottom

fix deepspeed model saving #370

Merged
merged 1 commit into from
May 18, 2022
Merged

fix deepspeed model saving #370

merged 1 commit into from
May 18, 2022

Conversation

pacman100
Copy link
Contributor

@pacman100 pacman100 commented May 18, 2022 •
edited

What does this PR do?

Fixes #369

  1. Fixes the deepspeed model saving in Zero Stage-3. This saves the model in 16-bit precision by first gathering all the parameters across processes onto main process to get complete state dict. A point to note - this is expensive in time and memory.

@pacman100 pacman100 requested a review from sgugger May 18, 2022 07:13
@pacman100 pacman100 mentioned this pull request May 18, 2022
Closed
@HuggingFaceDocBuilderDev
Copy link

HuggingFaceDocBuilderDev commented May 18, 2022 •
edited

The documentation is not available anymore as the PR was closed or merged.

sgugger
sgugger approved these changes May 18, 2022
View reviewed changes
Copy link
Collaborator

@sgugger sgugger 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 fixing! Should we have a minimal version of DeepSpeed? I don't think this was always supported.

@pacman100
Copy link
Contributor Author

Thanks for fixing! Should we have a minimal version of DeepSpeed? I don't think this was always supported.

Hello, yes, we should have minimal version of DeepSpeed. I will club it with all the other changes that I am planning to have to improve the DeeSpeed integration.

@pacman100 pacman100 merged commit d33dc39 into huggingface:main May 18, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@sgugger sgugger sgugger approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
3 participants
@pacman100 @HuggingFaceDocBuilderDev @sgugger