Re-enable colqwen 2.5 Omni; fix resize token embeddings#367
Re-enable colqwen 2.5 Omni; fix resize token embeddings#367ManuelFay merged 2 commits intoilluin-tech:mainfrom
Conversation
…n-tech#339 EDIT: that was based upon just looking at the main __init__.py. looking at the other files, perhaps it was intentionally removed...
|
@ManuelFay I did it as one PR, let me know if you'd prefer it split up. The commits are small and self contained. |
|
hahaha I love the HuggingFace sass ! |
ManuelFay
left a comment
ManuelFay
left a comment
There was a problem hiding this comment.
@QuentinJGMace Would be nice when you come back to have your inout on why this was commented out, but assuming it won't be very dangerous to comment this back in, I believe this can be merged
As an FYI, it certainly appears to be working fine! 😄 |
|
Yeah I looked at the removal PR and couldn't quite grasp the issue. But sometimes it's just "useless" to maintain or introduced too much discrepancies wrt to the rest of the codebase. |
) * 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
* Add ColQwen3 Support * fix(colqwen3): fix ruff lint errors * Re-enable colqwen 2.5 Omni; fix resize token embeddings (#367) * looks like colqwen 2.5 omni support was accidentally removed in #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 (#368) * lint * lint examples * Bump the transformer versions for ColQwen3 support * Fix lint error * Update performance number --------- Co-authored-by: tankm <kyemin.tan@tomoro.ai> Co-authored-by: Amrit Bath <TransAMrit@users.noreply.github.com> Co-authored-by: Manuel Faysse <43467008+ManuelFay@users.noreply.github.com>
2 participants
Fix issue #360 - for some reason, Omni looks to have been explicitly disabled, but not removed. Now you can use Omni again.
Additionally, get_token_embeddings() was broken as of the transformers upgrade. See the comments within that commit.