This is a mirror of the conceptual and coco image captioning weights from: https://github.com/rmokady/CLIP_prefix_caption
-