复兴 ConvNeXt：面向高效卷积扩散模型
Reviving ConvNeXt for Efficient Convolutional Diffusion Models
把 ConvNeXt 式的纯卷积骨干网络移植到条件扩散模型中，在 ImageNet 256/512 条件生成上以 DiT-XL/2 一半的浮点运算量达到相当性能，且训练步数少 7 倍以上；对做视觉生成的人而言，这是一条可直接替换 Transformer 骨干、显著降低训练成本的架构路线。
原文依据 using only 50% of the FLOPs of DiT-XL/2, FCDM-XL achieves competitive performance with 7x and 7.5x fewer training steps
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论近期的扩散模型日益青睐 Transformer 骨干网络，其动因是全注意力架构所展现出的卓越可扩展性。然而，局部性偏置、参数效率与硬件友好性——正是这些特性曾使卷积网络成为高效的视觉骨干——在现代生成建模中却鲜有探索。本文提出全卷积扩散模型（FCDM），其骨干网络与 ConvNeXt 类似，但面向条件扩散建模而设计。我们发现，FCDM-XL 仅使用 DiT-XL/2 一半的浮点运算量，在 256×256 与 512×512 分辨率下分别以少 7 倍和 7.5 倍的训练步数达到有竞争力的性能。值得注意的是，FCDM-XL 可在 4 块 GPU 的系统上完成训练，凸显了本架构出色的训练效率。结果表明，现代卷积设计为扩散模型的规模化提供了一种有竞争力且高度高效的替代方案，使 ConvNeXt 重新成为高效生成建模中简单而强大的构建模块。
英文原文摘要 1054 字符
Recent diffusion models increasingly favor Transformer backbones, motivated by the remarkable scalability of fully attentional architectures. Yet the locality bias, parameter efficiency, and hardware friendliness--the attributes that established ConvNets as the efficient vision backbone--have seen limited exploration in modern generative modeling. Here we introduce the fully convolutional diffusion model (FCDM), a model having a backbone similar to ConvNeXt, but designed for conditional diffusion modeling. We find that using only 50% of the FLOPs of DiT-XL/2, FCDM-XL achieves competitive performance with 7x and 7.5x fewer training steps at 256x256 and 512x512 resolutions, respectively. Remarkably, FCDM-XL can be trained on a 4-GPU system, highlighting the exceptional training efficiency of our architecture. Our results demonstrate that modern convolutional designs provide a competitive and highly efficient alternative for scaling diffusion models, reviving ConvNeXt as a simple yet powerful building block for efficient generative modeling.