Multimodal
#4
by TheBrinkster - opened
Could you quantize the multimodal version of this to GGUF for us?
Or, if it already works, which projector file do we use?
The original model available under is https://huggingface.co/wnma3mz/Janus-Pro-7B-LM is of architecture LlamaForCausalLM which is text only. If you link the multimodal version and it is supported by llama.cpp we will do so.
Is see you probably want https://huggingface.co/wnma3mz/Janus-Pro-7B. This is unfortunately of architecture JanusProConditionalGeneration which is not currently supported by llama.cpp.
Thanks for the reply. I learned something.
TheBrinkster changed discussion status to closed