Abstract: Diffusion Transformer (DiT), an emerging diffusion model for visual generation, has demonstrated superior perfor mance but suffers from substantial computational costs. Our investigations ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results