Duoia's picture
DuoVLM-40M v1: from-scratch 40M vision-language model (frozen CLIP + MiniPile-pretrained LM)
4afe981 verified