And this model along with it's config go in theComfyUI/custom_nodes/ComfyUI-champWrapper/checkpoints/image_encoder: https://huggingface.co/fudan-generative-ai/champ/tree/main/image_encoder Original repo: Champ: Controllable and Consistent Human Image Animation with 3D Parametric Guidance ...
raise AttributeError("ComfyUI version too old, can't autodetect properly. Set your dtype manually.") else: self.vae.to(convert_dtype(vae_dtype)) print(f"VAE using dtype: {self.vae.dtype}") pbar.update(1) # 2. unet converted_unet_config = create_unet_diffusers_config(original...