Closed AshD closed 5 months ago
@AshD Thanks, that is only possible if llama.cpp supports it - and currently it doesn't, unfortunately. I find that Phi-3-vision is actually really good, I hope they add support eventually.
Agreed. Someone opened an issue in llama.cpp for Phi-3 vision. https://github.com/ggerganov/llama.cpp/issues/7444
Would it be possible to get support for llava models. I think llama.cpp has support for those.
I don't think I currently have time to support that kind of work, but if they add support for phi-3-vision, then I would be quite interested in updating this codebase to support it.
Thanks @dranger003 I like the fact that this project is a thin wrapper around llama.cpp I am closing this issue now and will open a new one when the phi-3 vision support is added to llama.cpp
@AshD I created a quick API hosting in Python along with a C# client, if you want to check it out - while we're waiting for llama.cpp to support the model.
https://gist.github.com/dranger003/daff444ebf04951d4279b5b2dee71ab4
Thanks @dranger003 Looks good :-)
I do like the simplicity of this project's bindings to LLama.cpp.
Are there plans to add multimodal model support - like llava and phi-3 vision. I can test the bindings for these.
Thanks, Ash