There are many multimodal models that take an existing text only llm and do some visual training. Blip-2, Fromage, prismer and palm-e are examples. Multimodal i

There are many multimodal models that take an existing text only llm and do some visual training. Blip-2, Fromage, prismer and palm-e are examples. Multimodal isn't necessarily multimodality from scratch