Sunil Khatri

Hello. I have a clarification question regarding input grids and support pairs. It looks like the model always works with a 224×224 input image, which is tokenized into 14×14 patches....