Preprocessed image is not correct

Hi,
For the following cat image, I get this preprocessed input after the preprocessor. However, the preprocessed (after encoded by the ViT) looks not correct. Is this the correct behavior? Similar patterns happen for any kind of image for the model. Also, this applies to large, xl and xxl
```
cat_image = "/workspace/alpkant/cat.jpg"
preprocessed_example = preprocessor(text_inputs="", image_inputs=cat_image, text_targets= "Speed",target_modality="text", is_training=False)
plt.figure()
plt.imshow(preprocessed_example["/inputs/image/input"]*255) # This is because they are scaled between 0-1 after preprocessing
plt.show()
```