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

fix torch_dtype in estimate memory #3383

Merged
merged 4 commits into from
Feb 7, 2025
Merged

fix torch_dtype in estimate memory #3383

merged 4 commits into from
Feb 7, 2025

Conversation

SunMarc
Copy link
Member

@SunMarc SunMarc commented Feb 6, 2025

What does this PR do ?

Fixes #3379
This PR fixes the estimate memory command as it gave wrong results when torch_dtype is specified in the config we are loading.

@SunMarc SunMarc changed the title fix torch_dtype fix torch_dtype in estimate memory Feb 6, 2025
@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.

LGTM, thanks for the quick fix. Maybe let's add a test or at least a comment on why passing the torch dtype is needed.

@SunMarc
Copy link
Member Author

SunMarc commented Feb 6, 2025

Added a comment !

@BenjaminBossan
Copy link
Member

Linter is complaining, probably the line is too long.

@SunMarc SunMarc merged commit f19b957 into main Feb 7, 2025
28 checks passed
@SunMarc SunMarc deleted the fix-estimate-torch-dtype branch February 7, 2025 14:58
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.

estimate-memory gives ~64GB for meta-llama/Llama-3.1-70B-Instruct instead of anticipated ~140GB
3 participants