CompVis

/

stable-diffusion-v1-4

like 6.29k

Yeah, you need to accept conditions for using this model. Open the model page, accept the terms below and then go to your profile and create a new access token with read permissions. You're now able to clone the repository including the binary model files using git:

git lfs install
git clone "https://huggingface.co/CompVis/stable-diffusion-v1-4"
git lfs pull
git lfs checkout

When asked for credentials, as username enter hf_user and as password use the newly-created token. After all this is said and done, you can then import your model from the subdirectory:

# ...
pipe = StableDiffusionPipeline.from_pretrained("./stable-diffusion-v1-4", torch_dtype=torch.float16)
# ...

Note that there are alternative methods which don't use git but the huggingface_hub python package instead.

from diffusers import DiffusionPipeline

pipeline = DiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", use_auth_token=True)

Should work as well if you're logged in with:

huggingface-cli login

@patrickvonplaten

Using below code, I downloaded the model into my desktop. For my next usages, each run, do I need to download it all from beginning or is there any model assignment to pipe variable from my desktop? I think my model is downloaded into ./models path in my initial run. Am I right ?

pipe = StableDiffusionPipeline.from_pretrained(
"CompVis/stable-diffusion-v1-4",
torch_dtype=torch.float16,
scheduler=lms,
use_auth_token=True,
cache_dir=os.getenv("cache_dir", "./models")
).to(device)

Thank you,