Hi, I cannot find the scripts to construct the grounded_visual_chat_data for COCO datasets, can anyone find them? Thanks a lot!