Diffusion-RWKV is a diffusion model based on the RWKV architecture, designed to improve the scalability of diffusion models. It has been optimized and improved for image generation tasks, and can generate high-quality images. The model supports both unconditional and class-conditional training, with good performance and scalability.