Add LLaVA multi-modal support #789
Open
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
Enable packaging vision encoder (mmproj) and language model in a single llamafile. The vision encoder is now auto-detected, eliminating the need for --mmproj flag. Changes: - Allow multiple GGUF files in ZIP archives (llamafile.c) - Add auto-detection for common mmproj filenames (llava-cli.cpp) - Add llamafile_name() to get filename from handle - Include package_llava.sh helper script Example usage: ./package_llava.sh model.gguf mmproj.gguf llava.llamafile ./llava.llamafile --image photo.jpg -p "What's in this image?" 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <[email protected]>
- Detect when running as ./llava.llamafile executable - Set model path to executable when embedded - Allow mmproj auto-detection even when model path is empty
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Outdated suggestions cannot be applied.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.
LLaVA Multi-Modal Support
Makes it easier to use LLaVA models by packaging everything in one file.
Problem
LLaVA needs two files - the language model and vision encoder. This breaks the single-file philosophy.
Solution
Allow multiple GGUF files in one llamafile and auto-detect the vision encoder.
Changes
llamafile.c: Allow multiple GGUFs (just removed one error check)llava-cli.cpp: Detect when running as llamafile, auto-find mmprojpackage_llava.sh: Helper scriptUsage
Tested with mock files. Fixes TODO at line 180.