DPM-Solver-v3：具有實證模型統計的改進擴散ODE求解器

DPM-Solver-v3: Improved Diffusion ODE Solver with Empirical Model Statistics

October 20, 2023

作者: Kaiwen Zheng, Cheng Lu, Jianfei Chen, Jun Zhu

cs.AI

摘要

擴散概率模型（DPMs）在高保真度圖像生成方面表現出色，但採樣效率低下。最近的研究通過提出利用DPMs特定ODE形式的快速ODE求解器來加速採樣過程。然而，它們在推斷期間高度依賴特定的參數化（如噪聲/數據預測），這可能不是最佳選擇。在這項工作中，我們提出了一種新的配方，以實現在採樣期間最佳參數化，從而最小化ODE解的一階離散化誤差。基於這種配方，我們提出了DPM-Solver-v3，這是一種新的快速DPMs的ODE求解器，通過引入在預訓練模型上高效計算的幾個係數，我們稱之為經驗模型統計。我們進一步結合多步方法和預測校正框架，並提出一些技術，以改善在少量函數評估（NFE）或大導向尺度下的樣本質量。實驗表明，DPM-Solver-v3在無條件和有條件採樣中均取得了一致更好或可比的性能，無論是在像素空間還是潛在空間的DPMs中，尤其是在5至10 NFE。我們在無條件CIFAR10上實現了FID值為12.21（5 NFE）、2.51（10 NFE），在Stable Diffusion上實現了MSE值為0.55（5 NFE，7.5導向尺度），相對於先前的最新無需訓練方法，實現了15%至30%的加速。代碼可在https://github.com/thu-ml/DPM-Solver-v3找到。

English

Diffusion probabilistic models (DPMs) have exhibited excellent performance for high-fidelity image generation while suffering from inefficient sampling. Recent works accelerate the sampling procedure by proposing fast ODE solvers that leverage the specific ODE form of DPMs. However, they highly rely on specific parameterization during inference (such as noise/data prediction), which might not be the optimal choice. In this work, we propose a novel formulation towards the optimal parameterization during sampling that minimizes the first-order discretization error of the ODE solution. Based on such formulation, we propose DPM-Solver-v3, a new fast ODE solver for DPMs by introducing several coefficients efficiently computed on the pretrained model, which we call empirical model statistics. We further incorporate multistep methods and a predictor-corrector framework, and propose some techniques for improving sample quality at small numbers of function evaluations (NFE) or large guidance scales. Experiments show that DPM-Solver-v3 achieves consistently better or comparable performance in both unconditional and conditional sampling with both pixel-space and latent-space DPMs, especially in 5sim10 NFEs. We achieve FIDs of 12.21 (5 NFE), 2.51 (10 NFE) on unconditional CIFAR10, and MSE of 0.55 (5 NFE, 7.5 guidance scale) on Stable Diffusion, bringing a speed-up of 15\%sim30\% compared to previous state-of-the-art training-free methods. Code is available at https://github.com/thu-ml/DPM-Solver-v3.