Add model card
LLaVA-qwen25-3b-VLfull-r2-s42 s2b: inference weights + training history
initial commit