Does magma have a separate model for dealing with images?
#1
by
skyronhgs
- opened
So far I've used only gemma as a multimodal llm with quants on llama cpp, but besides that llava also seemed to always make available a CLIP model to use separatelly from the main model to deal with the image handling, parsing, conversion etc, I noticed there isn't one here, does perhaps magma include it all within the gguff? will it work directly with llama cpp for multimodal?
llama.cpp does not seem to support the model type MagmaForCausalLM for converting the safetensors to mmproj files at the moment.