Why '.png' is necessaary for label ? When I change the 'name_im_gt_list[0]["gt_ext"] = ".png" ' to ' name_im_gt_list[0]["gt_ext"] = ".jpg" ', an error occurred that "RuntimeError: The size of tensor a (4) must match the size of tensor b (3) at non-singleton dimension 0". Then I check my custom trainset with cv2.imread() and all shapes are [3000, 2000, 3] for image and [3000, 2000, 1] for gt, there is no image with shape[3000, 2000, 4]