Text-to-Image
stable-diffusion

How to get the weights!

#1
by cjlovering - opened

Hello! I’m sure it’s easy but I can’t see how to get the weights. I signed up and the email pointed me to this link but I don’t see how to get the files. Any pointers would be awesome!

Here you go: https://huggingface.co./CompVis/stable-diffusion-v-1-3/tree/main
Download the .ckpt file and perhaps look at this discussion for a Colab notebook that teaches you how to use them.

The public https://github.com/CompVis/stable-diffusion repo has instructions and example scripts if you are trying to run it locally (after you download weights above). Running great on a 3090 here.

This comment has been hidden

They are available to researchers, if you applied as a researcher and received an email inviting you, you should be able to see them there

@z8, please make sure you're are logged in locally when trying to download the weights. Before downloading the weights can you run

huggingface-cli login

in a shell? You can install huggingface-cli by running pip install huggingface_hub

Can anyone please provide me with the .ckpt, I applied but didn't receive a reply email.

@darkeye Be sure to check your junk mail folder. Lots of people have said that's where they have found messages from Stability AI.

This comment has been hidden
deleted
This comment has been hidden

Which one of you can share the weights file.thanks

I don't want to be that guy, but please refrain from begging for the weights. It's not a good look.

I'm sure that at this stage the developers don't want the weights to be freely distributed by the invited researchers. Otherwise, they would have made the weights public to begin with. Plus, it seems like the weights will be publicly released very soon (rumors are in the coming week). Have patience!

https://huggingface.co./CompVis/stable-diffusion-v-1-3/tree/main doesn't have a ckpt file, it has a bunch of directories (feature_extractor, vae, unet, etc) with .bin files. Am I misunderstanding something?

How to find model name, if I know only model hash, e.g. model hash 72506152?

Sign up or log in to comment