Official Codebase for POP-DIFFUSEQ: CONTROLLABLE SYMBOLIC MUSIC MULTI-INSTRUMENT INFILLING AND ACCOMPANIMENT GENERATION WITH LONG-AXIS ATTENTION The main function is ...
Official Codebase for DiffuSeq: Sequence to Sequence Text Generation With Diffusion Models. The diffusion process of our conditional diffusion language model DiffuSeq. It will take 2 more days to ...
Existing generative models, such as GAN, VAE, and flow-based models, suffer from mode collapse and must rely on surrogate objectives to approximate maximum likelihood training. Diffusion models have ...