Skip to content

How to reproduce your evals? #2

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
Open
eldarkurtic opened this issue Apr 16, 2025 · 2 comments
Open

How to reproduce your evals? #2

eldarkurtic opened this issue Apr 16, 2025 · 2 comments

Comments

@eldarkurtic
Copy link

Hey folks, great work on GPTQv2!

I would like to reproduce your W4g128 evaluations from the README.md, but I am getting significantly worse results than the ones you report here.

Image

Could you help me out to reproduce your evals by:

  1. releasing your model weights on HF-Hub or providing exact commands to recreate it (W4g128 GPTQv2)
  2. providing exact lm-evaluation-harness commands you used to obtain these scores
@yhhhli
Copy link
Contributor

yhhhli commented Apr 16, 2025 •
edited
Loading

The results were reported by the GPTQModel package, we recommend running the test files in their repository.

We will upload the checkpoints to Huggingface.

@Qubitium
Copy link

Qubitium commented Apr 17, 2025 •
edited
Loading

@eldarkurtic @yhhhli Please ref the GPTQModel issue (ModelCloud/GPTQModel#1545) for full reproduction of quant code, eval code, and HF uploaded models.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
3 participants
@Qubitium @eldarkurtic @yhhhli