Tools to train a generative model on arbitrary audio samples - pollinations/dance-diffusion