paper-with-me

Papers

FlashOmni: A Unified Sparse Attention Engine for Diffusion Transformers

2025-09-29 · Liang Qiao, Yue Dai, Yeqi Huang, Hongyu Kan, Jun Shi, Hong An arxiv

Multi-Modal Diffusion Transformers (DiTs) demonstrate exceptional capabilities in visual synthesis, yet their deployment remains constrained by substantial computational demands. To alleviate this bottleneck, many sparsity-based acceleration methods have been proposed. However, their diverse sparsity patterns often require customized kernels for high-performance inference, limiting universality. We propose FlashOmni, a unified sparse attention engine compatible with arbitrary DiT architectures. FlashOmni introduces flexible sparse symbols to standardize the representation of a wide range of sparsity strategies, such as feature caching and block-sparse skipping. This unified abstraction enables the execution of diverse sparse computations within a single attention kernel. In addition, FlashOmni designs optimized sparse GEMMs for attention blocks, leveraging sparse symbols to eliminate redundant computations and further improve efficiency. Experiments demonstrate that FlashOmni delivers near-linear, closely matching the sparsity ratio speedup (1:1) in attention and GEMM-$Q$, and achieves 2.5$\times$-3.8$\times$ acceleration in GEMM-$O$ (max peaking at about 87.5% of the theoretical limit). Applied with a multi-granularity sparsity strategy, it enables the Hunyuan model (33K) to achieve about 1.5$\times$ end-to-end acceleration without degrading visual quality.

📄 PDF Abstract BibTeX arXiv:2509.25401

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

SPADE: An Input-Adaptive Sparse Attention Engine for Fast Video Diffusion Models Inference

2026-08-04 · Shanghao Liu, Renze Chen, Size Zheng, Yuanqiang Liu 외 arxiv

Video diffusion transformers (vDiTs) generate high quality but pay quadratic self-attention cost, making inference prohibitive at video-token scales. The challenge is input-adaptive sparsity: selecting critical Q/K/V tok…

Video Generation

Fast Autoregressive Video Diffusion and World Models with Temporal Cache Compression and Sparse Attention

2026-02-02 · Dvir Samuel, Issar Tzachor, Matan Levy, Michael Green 외 arxiv

Autoregressive video diffusion models enable streaming generation, opening the door to long-form synthesis, video world models, and interactive neural game engines. However, their core attention layers become a major bot…

TurboDiffusion: Accelerating Video Diffusion Models by 100-200 Times

2025-12-18 · Jintao Zhang, Kaiwen Zheng, Kai Jiang, Haoxu Wang 외 arxiv

We introduce TurboDiffusion, a video generation acceleration framework that can speed up end-to-end diffusion generation by 100-200x while maintaining video quality. TurboDiffusion mainly relies on several components for…

Video Generation

QuantSparse: Comprehensively Compressing Video Diffusion Transformer with Model Quantization and Attention Sparsification

2025-09-28 · Weilun Feng, Chuanguang Yang, Haotong Qin, Mingqiang Wu 외 arxiv

Diffusion transformers exhibit remarkable video generation capability, yet their prohibitive computational and memory costs hinder practical deployment. Model quantization and attention sparsification are two promising d…

Video Generation

USV: Unified Sparsification for Accelerating Video Diffusion Models

2025-12-05 · Xinjian Wu, Hongmei Wang, Yuan Zhou, Qinglin Lu arxiv

The scalability of high-fidelity video diffusion models (VDMs) is constrained by two key sources of redundancy: the quadratic complexity of global spatio-temporal attention and the computational overhead of long iterativ…

Video Generation