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

[BugFix] Fix Marlin24 Bug #203

Merged
merged 1 commit into from
Nov 8, 2024
Merged

[BugFix] Fix Marlin24 Bug #203

merged 1 commit into from
Nov 8, 2024

Conversation

dsikka
Copy link
Contributor

@dsikka dsikka commented Nov 8, 2024

Summary

  • Fix the check for permutations which have been incorrectly been running for the channel-wise case
  • Idk how this ever worked for group quant

Generations Using Group Quantization in vLLM:

[' Paris and it is located in the north-west of the country.\nThe population of', ' the leader of the executive branch of the federal government, and thus the head of', ' Liz. I live in a small town in the middle of a prairie.']

Copy link
Member

@rahul-tuli rahul-tuli left a comment

Choose a reason for hiding this comment

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

Whaaaaaaaaaaaaaaaaaaat 🥲

Copy link
Contributor

@kylesayrs kylesayrs left a comment

Choose a reason for hiding this comment

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

This should just be a syntax warning? Not sure if I see why this would raise an error

@rahul-tuli
Copy link
Member

This should just be a syntax warning? Not sure if I see why this would raise an error

This doesn't raise an error, but before this diff the group quantized models with marlin_24 compressor were giving 0 accuracy

@rahul-tuli rahul-tuli merged commit db6ccb2 into main Nov 8, 2024
1 check passed
@rahul-tuli rahul-tuli deleted the fix_marlin24 branch November 8, 2024 03:29
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.

3 participants