ZYI Turbo 0.1
Small experimental general-purpose text-to-image diffusion denoiser.
Model
- Trainable parameters: 18.77M
- Resolution: 128x128
- Architecture: Diffusers
UNet2DConditionModel
- Text encoder:
openai/clip-vit-base-patch32
- VAE:
stabilityai/sd-vae-ft-mse
- Recommended inference steps: 12
- Recommended CFG: 5.0
Important
This repository contains the trained ZYI Turbo denoiser.
The CLIP text encoder and VAE are not necessarily bundled with the model and may need to be loaded separately during inference.
