GLIDE: a diffusion-based text-conditional image synthesis model - View it on GitHub