Add ColQwen3 Support #366
Open
Add ColQwen3 Support #366
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Collaborator
|
Some tests don't run. (2) The tests. should we bump the transformers version and we're good ? |
Contributor
|
@hxssgaa thanks for bringing these over. Having a blast using them! |
Author
We have fixed ruff lint error in our commits, the other lint errors are not from us. For the transformers version it must be at least |
|
I try to run tomoro-colqwen3-embed-8b with this PR, and I encounter following errors: How to fix it? or Is it my wrong configuration? |
Author
Hi, note that Colpali format isn't directly compatible with current Tomoro-colqwen3 hf models as they are converted from this Colpali format, you can refer to the Tomoro HF repo for how to run the models for now. We intent to also share the conversion script later, but it shouldn't belong to this repo. |
Collaborator
|
So.one thing I don't understand yet is that this PR is not compatible with the Tomoro checkpoints you shared ? Isn't it better to make this compatible ? |
|
We based our huggingface repo off the colqwen2 implementation on transformers here: https://github.com/huggingface/transformers/tree/main/src/transformers/models/colqwen2, which uses a different naming convention for params compared to this repo. Should we unify the names or keep 2 separate versions? |
|
I trained a 2B smaller colqwen3 model on the Qwen3-VL-2B-Instruct model, welcome to follow and use~ |
ManuelFay
requested changes
Dec 26, 2025
Collaborator
ManuelFay
left a comment
ManuelFay
left a comment
There was a problem hiding this comment.
Is it possible to rebase on main so that linting gets corrected ?
The pther interrogation I have is whether the pyprojecxt should get updatred. From my understanding, qwen3 is only in newer versions of transformers so I am guessing we should bump the minimal transformers package right ?
The rest looks great !
) * looks like colqwen 2.5 omni support was accidentally removed in illuin-tech#339 EDIT: that was based upon just looking at the main __init__.py. looking at the other files, perhaps it was intentionally removed... * found & fixed resize_token_embeddings() breakage
* lint * lint examples
Author
Already rebased to main, and yes, I have updated the minima transformer version to be |
ManuelFay
requested changes
Dec 26, 2025
tests/models/modernvbert/test_interpretability_colmodernvbert.py
Outdated
Show resolved
Hide resolved
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
7 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Add ColQwen3 support
The fine-tuned colqwen3 models are below including with benchmark results: