Light Forcing: Accelerating Autoregressive Video Diffusion via Sparse Attention

TL;DR

Light Forcing introduces a novel sparse attention mechanism tailored for autoregressive video generation, significantly improving efficiency and quality while maintaining high fidelity in generated videos.

Contribution

It presents the first sparse attention solution specifically designed for AR video models, incorporating Chunk-Aware Growth and Hierarchical Sparse Attention for better context utilization.

Findings

Outperforms existing sparse attention in quality (84.5 on VBench)

Achieves 1.2-1.3x speedup in end-to-end inference

Combined with FP8 quantization, reaches 19.7 FPS on RTX 5090

Abstract

Advanced autoregressive (AR) video generation models have improved visual fidelity and interactivity, but the quadratic complexity of attention remains a primary bottleneck for efficient deployment. While existing sparse attention solutions have shown promise on bidirectional models, we identify that applying these solutions to AR models leads to considerable performance degradation for two reasons: isolated consideration of chunk generation and insufficient utilization of past informative context. Motivated by these observations, we propose \textsc{Light Forcing}, the \textit{first} sparse attention solution tailored for AR video generation models. It incorporates a \textit{Chunk-Aware Growth} mechanism to quantitatively estimate the contribution of each chunk, which determines their sparsity allocation. This progressive sparsity increase strategy enables the current chunk to inherit prior knowledge in earlier chunks during generation. Additionally, we introduce a \textit{Hierarchical Sparse Attention} to capture informative historical and local context in a coarse-to-fine manner. Such two-level mask selection strategy (\ie, frame and block level) can adaptively handle diverse attention patterns. Extensive experiments demonstrate that our method outperforms existing sparse attention in quality (\eg, 84.5 on VBench) and efficiency (\eg, $1.2{\sim}1.3\times$ end-to-end speedup). Combined with FP8 quantization and LightVAE, \textsc{Light Forcing} further achieves a $2.3\times$ speedup and 19.7\,FPS on an RTX~5090 GPU. Code will be released at \href{https://github.com/chengtao-lv/LightForcing}{https://github.com/chengtao-lv/LightForcing}.