多解析度流匹配：透過分階段採樣實現無需訓練的擴散加速

摘要

與硬體無關的文字轉圖像擴散加速策略，例如時間步蒸餾與特徵快取，可在無需客製化內核或系統層級最佳化的情況下縮短推論時間。其中，多解析度生成策略近期獲得廣泛關注，在不經任何訓練的情況下實現超過5倍加速。然而，在潛空間中執行上取樣，並搭配局部區域的選擇性修改，導致這些方法出現明顯的模糊或偽影。為此，我們提出MrFlow，一種基於分階段低解析度到高解析度管線的訓練免除多解析度加速策略，適用於預訓練的流匹配模型。MrFlow首先在低解析度下快速生成主要結構，接著使用輕量級的預訓練GAN模型在像素空間中進行超解析度，隨後注入低強度雜訊以實現高頻重取樣，最後在高解析度下優化細節。在FLUX.1-dev與Qwen-Image上的定量與定性結果顯示，MrFlow利用低解析度取樣的二次性令牌減少與步驟需求降低，實現端到端10倍加速，同時在OneIG指標上與加速前差距維持在1%以內，顯著超越其他免訓練加速策略，且完全無需訓練或運行時動態識別。MrFlow更可進一步與預先訓練的時間步蒸餾策略正交直接結合，實現高達25倍的更高生成加速。

English

Hardware-agnostic strategies for accelerating text-to-image diffusion, such as timestep distillation and feature caching, can reduce inference time without custom kernels or system-level optimization. Among them, multi-resolution generation strategies have recently received broad attention, attaining more than 5x speedup without any training. However, the design of performing upsampling in the latent space, together with the selective modification of partial regions, causes these methods to exhibit noticeable blurring or artifacts. To this end, we propose MrFlow, a training-free multi-resolution acceleration strategy for pretrained flow-matching models built upon a staged low-to-high-resolution pipeline. MrFlow first rapidly generates the main structure at low resolution, then performs super-resolution in the pixel space using a lightweight pretrained GAN-based model, subsequently injects low-strength noise to enable high-frequency resampling, and finally refines the details at high resolution. Quantitative and qualitative results on FLUX.1-dev and Qwen-Image show that MrFlow exploits the quadratic token reduction and reduced step requirement of low-resolution sampling to achieve 10x end-to-end acceleration while keeping OneIG within a 1% gap relative to that before acceleration, significantly surpassing other training-free acceleration strategies, and requiring no training or runtime dynamic identification whatsoever. MrFlow can further be directly combined orthogonally with pre-trained timestep distillation strategies, achieving even higher generation acceleration of up to 25x.