Llama3.2-11b-Vision

[gabrielle-ong]

No description provided.

[hahuyhoang411]

llama.cpp state: ggerganov/llama.cpp#9643

[dan-menlo]

@hahuyhoang411 @vansangpfiev I'd like to take next sprint to see if we can contribute Vision back to llama.cpp.

I think a good starting point is to build on top of @ngxson's PR here: ggerganov/llama.cpp#9687

We need an extensible architecture on top of llama.cpp for multi-modality. For Ichigo, we likely need to generate audio tokens in a separate Python process though (i.e. whisper-speech)