# make sure you're logged in with `huggingface-cli login` from torch import autocast from diffusers import StableDiffusionPipeline pipe = StableDiffusionPipeline.from_pretrained( "CompVis/stable-diffusion-v1-4", use_auth_token=True ).to("cuda") prompt = "a photo of an astronaut riding a horse on mars" with autocast("cuda"): image = pipe(prompt)["sample"][0] image.save("astronaut_rides_horse.png") #814