How to inference the image generation instead of editing?

[xingzhejun]

Hi!

Such a nice work! I am following up these line of work. May I ask the image generation code? Notice that in inference.py there is only the editing code.

[xingzhejun]

I tried the generation modifying the editing code. The result is a bullshit.

[xingzhejun]

I used the example in the paper: "A trio of chairs, a trio of banners, a trio of desks, a quartet of ships, and a quartet of cushions.". Please see the awful work done by this model!