metadata
language:
- en
tags:
- stable-diffusion
- text-to-image
license: bigscience-bloom-rail-1.0
inference: false
copy pasted from the genshin stable inversion repo, all you need to know is that it's the token_embedding layer and you can load it in there with torch.load
A finetuning equivalent with less VRAM requirement than finetuning Stable Diffusion itself, faster if you have all the images downloaded, less space taken up by the models since you only need CLIP
A notebook for producing your own "stable inversions" is included in this repo but I wouldn't recommend doing so (they suck). It works on Colab free tier though.
link to notebook for you to download
how you can load this into a diffusers-based notebook like Doohickey might look something like this (later edit: this is now included in Doohickey BETA by default)
from huggingface_hub import hf_hub_download
stable_inversion = "user/my-stable-inversion" #@param {type:"string"}
inversion_path = hf_hub_download(repo_id=stable_inversion, filename="token_embeddings.pt")
text_encoder.text_model.embeddings.token_embedding.weight = torch.load(inversion_path)