Skip to content

fix: update colpali engine models #2905

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
Merged
merged 20 commits into from
Jul 16, 2025
Merged

fix: update colpali engine models #2905

merged 20 commits into from
Jul 16, 2025

Conversation

paultltc
Copy link
Contributor

@paultltc paultltc commented Jul 15, 2025 •
edited
Loading

  • fixes Cannot pass in revision in colpali engine #2851
  • add tqdm to embedding functions for better logging
  • adapt to new process_text function from colpali engine (replacing process_query)
  • computes similarity scores based on model's processor (will allow to add engine BiEnc models later)
    @isaac-chung

@paultltc paultltc changed the title fix: pass revision to colpali engine models fix: update colpali engine models Jul 15, 2025
@Samoed Samoed requested a review from isaac-chung July 15, 2025 22:30
isaac-chung
isaac-chung approved these changes Jul 16, 2025
View reviewed changes
Copy link
Collaborator

@isaac-chung isaac-chung left a comment

Choose a reason for hiding this comment

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

Thanks @paultltc ! Just to confirm the scores are the same, would you mind running the same model on 1 dataset on this PR and on the main branch to compare?

PR main
dataset1 score1 score 2

@paultltc
Copy link
Contributor Author

paultltc commented Jul 16, 2025 •
edited
Loading

I fixed 2 main things:

  • colqwen2.5 had an error in its revision
  • adjusted the number of query augmentation tokens to match previous implementation of process_queries

Also, the version of colpali_engine is updated to match latest transformers version changes.

PR vs Main Branch comparison

PR main
VidoreArxivQARetrieval 0.88868 0.88868

NB: I checked the scores on all splits of ViDoRe(v1) and it matches perfectly

Current Leaderboard Scores

The scores vary slightly with the current ones on the leaderboard (might be due to changes in our package or in transformers). Also some revisions have changed. I will open a PR on the results repo to update these!

@isaac-chung @ManuelFay

@paultltc
Copy link
Contributor Author

plus I am double checking all revisions to avoid other copy paste errors lol

@paultltc
Copy link
Contributor Author

Tested them all all good! Benchmarks are running

This was referenced Jul 16, 2025
Closed
Merged
@isaac-chung isaac-chung merged commit 9864e2a into embeddings-benchmark:main Jul 16, 2025
10 checks passed
@paultltc paultltc deleted the fix/revision branch July 16, 2025 19:15
@isaac-chung
Copy link
Collaborator

@Samoed FYI for when we merge main into v2 branch next.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@isaac-chung isaac-chung isaac-chung approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

Cannot pass in revision in colpali engine
2 participants
@paultltc @isaac-chung