8000 Gemma3 by vince62s · Pull Request #234 · eole-nlp/eole · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

Gemma3 #234

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

Merged
merged 17 commits into from
Apr 25, 2025
Merged

Gemma3 #234

merged 17 commits into from
Apr 25, 2025

Conversation

vince62s
Copy link
Contributor
@vince62s vince62s commented Apr 22, 2025

Pushing a cleaner PR with changes from Main only.

It works fine for both Llava-type (pixtral/Mistral-24b) and Gemma3.

However still pending but maybe for later PR:

  • clean the image_utils.py tools
  • revamp the vision.py VisionEncoder ? (for better split between llava vs Gemma)

Open questions:
It seems that image masking (required for Gemma3?) is still ok for Llava (sounds reasonable). Hence I left the "decoder_in" arg in self.model.decoder() call from inference.py

@francoishernandez francoishernandez mentioned this pull request Apr 22, 2025
Copy link
Member
@francoishernandez francoishernandez left a comment

Choose a reason for hiding this comment

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

Some comments/discussion topics.

Additional notes, can be reviewed in other PR(s):

  • some new config fields lack helpstrings;
  • image_utils is a bit messy, we can probably strip it down at some point.

@vince62s vince62s merged commit 9e559f3 into eole-nlp:main Apr 25, 2025
2 checks passed
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.

2 participants
0