Looking for GGUF

#2
by ptrkzr - opened

Maybe @bartowski or @MaziyarPanahi may help?

New architecture that's not supported in llama.cpp yet :')

@bartowski thanks for the info. I thought this model is based on Llama3 8B, but you're right, there's another "HeteroGraphLLMForCausalLM" architecture in config.json.

So maybe @GraphAgent can shed some light on this.

This comment has been hidden

Thanks for the discussion. We will consider implement it with llama.cpp with reference to that of Llava, if any.
It is worth mention that we have basically very similar architecture with Llava, code reference: https://github.com/HKUDS/GraphAgent/blob/main/GraphAgent-inference/graph_action_agent/graphllm/graphllm.py

We would appreciate any forms of contribution! :)

Sign up or log in to comment