Hugging Face's logo

CompVis
/
stable-diffusion-v1-4

Text-to-Image
Diffusers
Safetensors
StableDiffusionPipeline
stable-diffusion
stable-diffusion-diffusers
Model card Files
xet
 Community
280

OSError: /CompVis/stable-diffusion-v1-4/ does not appear to have a file named config.json.

#237
by Kayell - opened
Discussion
Kayell

I downloaded the model file to my local directory and loaded it from there, throwing this error“ /CompVis/stable-diffusion-v1-4-fp16/ does not appear to have a file named config.json”. My transformers library is the latest version and my code is as follows：

HfFolder.save_token("xxxx")
model_path = "/xxx/"
config = AutoConfig.from_pretrained(model_path)
pipe = StableDiffusionPipeline.from_pretrained(model_path, use_auth_token=True,config=config)

Can anyone tell me how to solve this issue?
Kayell changed discussion title from OSError: /mypath/CompVis/stable-diffusion-v1-4-fp16/ does not appear to have a file named config.json. to OSError: /CompVis/stable-diffusion-v1-4/ does not appear to have a file named config.json.
pcuenq
 CompVis org

Hi @Kayell , you don't need to use the config, and you don't need the token either if the files are locally in your filesystem. The following should work:

model_path = "/xxx/"
pipe = StableDiffusionPipeline.from_pretrained(model_path)
Kayell
This comment has been hidden
Kayell

@pcuenq Hello！Thank you for your response！ I made the changes to my code as you suggested, but I'm still encountering the same error when running it. Can you help me?
kafai-chan
This comment has been hidden
Kayell

@kafai-chan Hey! I can't see your comment because it's hidden. Could you tell me what you replied?
JaneWen

@Kayell Hey! I have the same issue. I have researched the issue online and someone said it can be solved by updating the transformers to the latest version. But I have the version of 4.36.0 and it still happened. Do you solve the question? If you've worked it out, could you please tell me the solution?
Kayell

@JaneWen Sorry, I still have no idea. Maybe we can explore this issue together
JaneWen

Hi, I have solved the issue. The problem is I didn't log in to the huggingface. I use the command huggingface-cli login and the pretrained model can be downloaded successfully.
Kayell

@JaneWen Congratulations! There is no need to login because I have downloaded the model file to my local directory. I'm not sure what caused this error. I change the version to stablediffusion2, and it works well.

Sign up or log in to comment