Small-Multimodal-Vision-Model Small Multimodal Vision Model "Imp-v1-3b" trained using Phi-2 and Siglip.