通过草稿-精修并行化扩散模型采样，显著提速且保持生成质量。
DRiffusion: Draft-and-Refine Process Parallelizes Diffusion Models with Ease
- 用跳步生成多个未来时间步的草稿状态并行计算噪声。
- 实测速度提升1.4倍至3.7倍，质量损失极小。
- 适合对延迟敏感的交互式生成应用。
扩散模型在高保真内容生成中表现卓越，但采样过程迭代缓慢，导致延迟高，限制了其在交互式应用中的使用。我们提出 DRiffusion，一种通过草稿-精修流程并行化扩散推理的框架。该方法利用跳步机制生成未来多个时间步的草稿状态，并行计算其对应噪声，再用于标准去噪过程生成优化结果。理论上，该方法在保守或激进模式下分别实现 $\tfrac{1}{n}$ 或 $\tfrac{2}{n+1}$ 的加速比，其中 $n$ 表示设备数。实验表明，DRiffusion 在多个扩散模型上实现 1.4×–3.7× 的加速，同时生成质量几乎不受影响：在 MS-COCO 数据集上，FID 和 CLIP 指标与原模型基本一致，PickScore 与 HPSv2.1 平均仅下降 0.17 和 0.43。结果验证了该方法能显著加速并维持感知质量。
原文摘要 · Abstract (English)
Diffusion models have achieved remarkable success in generating high-fidelity content but suffer from slow, iterative sampling, resulting in high latency that limits their use in interactive applications. We introduce DRiffusion, a parallel sampling framework that parallelizes diffusion inference through a draft-and-refine process. DRiffusion employs skip transitions to generate multiple draft states for future timesteps and computes their corresponding noises in parallel, which are then used in the standard denoising process to produce refined results. Theoretically, our method achieves an acceleration rate of $\tfrac{1}{n}$ or $\tfrac{2}{n+1}$, depending on whether the conservative or aggressive mode is used, where $n$ denotes the number of devices. Empirically, DRiffusion attains 1.4$\times$-3.7$\times$ speedup across multiple diffusion models while incur minimal degradation in generation quality: on MS-COCO dataset, both FID and CLIP remain largely on par with those of the original model, while PickScore and HPSv2.1 show only minor average drops of 0.17 and 0.43, respectively. These results verify that DRiffusion delivers substantial acceleration and preserves perceptual quality.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。