無需分類器自由引導的擴散模型

Diffusion Models without Classifier-free Guidance

February 17, 2025

作者: Zhicong Tang, Jianmin Bao, Dong Chen, Baining Guo

cs.AI

摘要

本文提出了一種新穎的訓練目標——模型引導（Model-guidance, MG），旨在解決並取代目前廣泛使用的無分類器引導（Classifier-free guidance, CFG）。我們的創新方法超越了僅對數據分佈進行建模的標準做法，轉而將條件後驗概率納入考量。該技術源自CFG的理念，既簡便又高效，可作為即插即用模塊應用於現有模型。我們的方法顯著加速了訓練過程，使推理速度翻倍，並達到了與甚至超越當前採用CFG的擴散模型相媲美的卓越質量。大量實驗證明了該方法在不同模型和數據集上的有效性、效率及可擴展性。最終，我們在ImageNet 256基準測試中取得了1.34的FID值，創下了最新記錄。代碼已開源於https://github.com/tzco/Diffusion-wo-CFG。

English

This paper presents Model-guidance (MG), a novel objective for training diffusion model that addresses and removes of the commonly used Classifier-free guidance (CFG). Our innovative approach transcends the standard modeling of solely data distribution to incorporating the posterior probability of conditions. The proposed technique originates from the idea of CFG and is easy yet effective, making it a plug-and-play module for existing models. Our method significantly accelerates the training process, doubles the inference speed, and achieve exceptional quality that parallel and even surpass concurrent diffusion models with CFG. Extensive experiments demonstrate the effectiveness, efficiency, scalability on different models and datasets. Finally, we establish state-of-the-art performance on ImageNet 256 benchmarks with an FID of 1.34. Our code is available at https://github.com/tzco/Diffusion-wo-CFG.