runwayml

/

stable-diffusion-inpainting

like 1.46k

running with on personal develop machine with gpu, but get wired result
run1:
pipe = StableDiffusionInpaintPipeline.from_pretrained(
"models/runwayml/stable-diffusion-inpainting",
)
prompt = "Face of a yellow cat, high resolution, sitting on a park bench"
image = pipe(prompt=prompt, image=image,
mask_image=mask_image).images[0]
image.save("./outputs/inpainting/yellow_cat_on_park_bench.png")

input image:
https://huggingface.co/datasets/emma7033/result_inpainting/blob/main/dog.png
output:
https://huggingface.co/datasets/emma7033/result_inpainting/blob/main/yellow_cat_on_park_bench.png

run2:
pipe = StableDiffusionInpaintPipeline.from_pretrained(
"models/runwayml/stable-diffusion-inpainting",
revision="fp16",
)
https://huggingface.co/datasets/emma7033/result_inpainting/blob/main/yellow_cat_on_park_bench_fp16.png

run3:
pipe = StableDiffusionInpaintPipeline.from_pretrained(
"models/runwayml/stable-diffusion-inpainting",
revision="fp16",
torch_dtype=torch.float16,
)
"LayerNormKernelImpl" not implemented for 'Half'