Hi, Nice work! Does this work for pre-trained ImageNet1k ViTs ? or rather models just trained for image classification ? Best, Sid