Bare bones Colab notebook for StableDiffusion
I struggled to get the first Colab notebook for StableDiffusion working, and I wanted a really bare-bones version, so I went back to the original model card and put together a very simple 15-line Colab which I'm going to use as the basis for further experiments. It's available here, if anyone wants to try it out:
https://colab.research.google.com/drive/1BLWHIP3yeuYO2dEt5wXhs8-itKCOI-rW?usp=sharing
You'll need your Hugging Face user id and access token to use it.
Thanks for sharing this! I can verify that this works with an A100.
This isn't working for me, I'm getting the error:
{'bert', 'vqvae'} was not found in config. Values will be initialized to default values.
TypeError Traceback (most recent call last)
/tmp/ipykernel_25398/4014842543.py in
3 from diffusers import LDMTextToImagePipeline
4 import numpy as np
----> 5 pipe = LDMTextToImagePipeline.from_pretrained("stable-diffusion-v1-3-diffusers")
/opt/conda/lib/python3.7/site-packages/diffusers/pipeline_utils.py in from_pretrained(cls, pretrained_model_name_or_path, **kwargs)
189
190 # 5. Instantiate the pipeline
--> 191 model = pipeline_class(**init_kwargs)
192 return model
193
TypeError: init() missing 2 required positional arguments: 'vqvae' and 'bert'
Check out the updated code at the repo for the diffusers: https://huggingface.co/CompVis/stable-diffusion-v1-3-diffusers
All you need to add besides the example code is your Huggingface token
{'bert', 'vqvae'} was not found in config. Values will be initialized to default values.
Hi
@neelmango
, can you please try StableDiffusionPipeline instead of LDMTextToImagePipeline? A new version of Diffusers is coming out soon and these details will be ironed out and clarified :)
hm. didn't get that hint, imho sth went wrong while renaming stuff
works for me tho if i fix the commit with
%cd stable-diffusion-v1-3-diffusers
!git checkout e886cdb9fb59db7bd208a4e87665532ff27da41c
%cd ..
{'bert', 'vqvae'} was not found in config. Values will be initialized to default values.
Hi @neelmango , can you please try
StableDiffusionPipelineinstead ofLDMTextToImagePipeline? A new version of Diffusers is coming out soon and these details will be ironed out and clarified :)
results in
TypeError: init() missing 2 required positional arguments: 'vae' and 'text_encoder'
Could you try the colab embedded in
https://huggingface.co/blog/stable_diffusion