metadata
datasets:
- irodkin/celeba_with_llava_shorter_captions
For this model the only one available contol input is the conditioning_image from dataset irodkin/celeba_with_llava_shorter_captions
And this image the model actually ignores. Its just experimental text2image