Any plan to support UniDiffuser? https://github.com/thu-ml/unidiffuser versatile diffusion(VD) maybe won't satisfy for multimodal