clip.cpp / gguf-py: Support for Qwen2.5 VL - WIP / REVIEW NEEDED (#11483) #12119
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
4 tasks
vladislavdonchev
changed the title
vladislavdonchev
changed the title
vladislavdonchev
changed the title
|
This will be reopened as a follow-up to the vision API refactor: #11292 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
Fellas, I gotta say - I grossly underestimated this task...
Work is still in progress in our fork here (comment here if you want to review the code, so we can update the PR):
https://github.com/Independent-AI-Labs/llama.cpp/commits/debug/build/
Qwen2.5 is a total overhaul of the Qwen VLM and introduces a bunch of concepts that require special handling.
I've documented my approach in this paper here:
https://github.com/Independent-AI-Labs/local-super-agents/blob/main/res/docs/papers/Implementing%20Qwen2.5VL.pdf
Comments & ideas are welcome!