LLaMa-3-8B mmproj?

#2
by Joseph717171 - opened

We need to give LLaMa-3-8B vision. Can you help? πŸ™

Koboldcpp org

I can only add it if a proper Llama 3 8b llava finetune is created.

Koboldcpp org

Its been added.

https://huggingface.co/xtuner/llava-llama-3-8b-v1_1

I'm trying to run this with the newly added mmproj adapter in LMStudio, quantized as Q5_K_M, but performance isn't great.. has anybody experimented with it already?
What are your experiences with it?

Sign up or log in to comment