This is an implementation of a Cog model. Cog packages machine learning models as standard containers.
First, download the pre-trained weights:
cog run script/download-weights
Then, you can run
text2img predictions:
cog predict -i prompt="your prompt"
img2img predictions:
cog predict -i image=@astro.png -i prompt="photo of a lone astronaut standing on a barren planet"
inpainting predictions:
cog predict -i image=@demo.png -i mask=@mask.png -i prompt="a tabby cat, high resolution, sitting on a park bench"