README.md

---
license: openrail
---

VAE for anime style checkpoints

Converted pastel-waifu-diffusion.vae.pt in https://huggingface.co./andite/pastel-mix to diffusers format. 

Example usage
```
from diffusers import AutoencoderKL, StableDiffusionPipeline
import torch

pipe = StableDiffusionPipeline.from_pretrained(
    pretrained_model_name_or_path='andite/anything-v4.0',
    vae = AutoencoderKL.from_pretrained('lint/anime_vae')
)

for component in pipe.components.values():
    if hasattr(component, 'device'):
        component.to('cuda', torch.float16)

out = pipe('1girl, blue eyes')
out.images[0]

```