Question at inference stage

[Kris-Lcq]

Hi, authors. You guys work are so amazing. Thank you for your efforts for safe AIGC.
I have one simple question at inference stage. The white model used in this repository is "/runwayml/stable-diffusion-inpainting"，if i only use the text-prompt attack to generate prompt, how can i generate an image with just the prompt? Because，the official hugging face repository of say "stable-diffusion-inpainting" needs three inputs, prompt、image and masked image whereas i only have the optimaized prompt.

[yangyijune]

Thanks for your question. The text encoders used in both SD-inpainting and SD are the same. To conduct prompt attacks, you can change the repository path to runwayml/stable-diffusion-v1-5 and use the corresponding pipeline for image generation without a mask and the original image as inputs.