3 ms·
they use autoencoder/autoregressive model to predict the big picture, and diffusion for the details, similar to yours. The difference is they use discrete toke
by billconan 2y ago
they use autoencoder/autoregressive model to predict the big picture, and diffusion for the details, similar to yours.
The difference is they use discrete tokens.