HackerPapers

arXiv:2212.09748cs.CV

Scalable Diffusion Models with Transformers

William Peebles, Saining Xie

Abstract

We explore a new class of diffusion models based on the transformer architecture. We train latent diffusion models of images, replacing the commonly-used U-Net backbone with a transformer that operates on latent patches. We analyze the scalability of our Diffusion Transformers (DiTs) through the lens of forward pass complexity as measured by Gflops. We find that DiTs with higher Gflops -- through increased transformer depth/width or increased number of input tokens -- consistently have lower FID. In addition to possessing good scalability properties, our largest DiT-XL/2 models outperform all prior diffusion models on the class-conditional ImageNet 512x512 and 256x256 benchmarks, achieving a state-of-the-art FID of 2.27 on the latter.

arXiv comments
Code, project page and videos available at https://www.wpeebles.com/DiT
Journal reference
DOI
Subjects
cs.CVComputer Vision and Pattern Recognition cs.LGMachine Learning

On Hacker News

Linked by 4 identifiable commenters in 4 comments across 4 threads, from Feb 2024 to Jun 2025. Found in 1 HN submission from Aug 2023 to Aug 2023. Open submission discussions.

Comments linking this paper, per year.

Discussed in comments4 threads

  1. Normalizing Flows Are Capable Generative Models 1 comment

    • It's a bit more complicated than that and I don't think you're being fair. StarFlow and the AR models are fixed but DiT is being compared at different amount of steps and we don't really care if we generate garbage at blazing speeds[0]. Go look at... also Figure 10 (lol) from…

      godelskiView on HNParent context

  2. Gemini Diffusion 1 comment

    • Image diffusion models also use transformers nowadays. Here is the original "diffusion transformer" paper: https://arxiv.org/abs/2212.09748 Earlier image diffusion models used U-nets: https://en.wikipedia.org/wiki/U-Net

      cubefoxView on HNParent context

  3. Stable Diffusion 3 1 comment

    • I understand that Sora is very popular, so it makes sense to refer to it, but when saying it is similar to Sora, I guess it actually makes more sense to say that it uses a Diffusion Transformer (DiT) (https://arxiv.org/abs/2212.09748) like Sora. We don't really know more details…

      albertzeyerView on HNParent context

  4. Sora: Creating video from text 1 comment

    • On the other hand, this is the first convincing use of a “diffusion transformer” [1]. My understanding is that videos and images are tokenized into patches, through a process that compresses the video/images into abstracted concepts in latent space. Those patches (image/video…

      dr_dshivView on HNParent context

Submitted to Hacker News

Connected through shared HN threads, not inferred similarity or a recommendation.