A library of pretrained diffusion models across multiple modalities, such as vision and audio.