HilbertA: Hilbert Attention for Image Generation with Diffusion Models

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Zheng, Shaoyi, Lu, Wenbo, Xia, Yuxuan, Liu, Haomin, Wang, Shengjie
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866915525741772800
author Zheng, Shaoyi
Lu, Wenbo
Xia, Yuxuan
Liu, Haomin
Wang, Shengjie
author_facet Zheng, Shaoyi
Lu, Wenbo
Xia, Yuxuan
Liu, Haomin
Wang, Shengjie
contents Designing sparse attention for diffusion transformers requires reconciling two-dimensional spatial locality with GPU efficiency, a trade-off that current methods struggle to achieve. Existing approaches enforce two-dimensional spatial locality but often incur uncoalesced memory access. We present HilbertA, a 2D-aware and GPU-efficient sparse attention mechanism. HilbertA reorders image tokens along Hilbert curves to achieve a contiguous memory layout while preserving spatial neighborhoods, and employs a sliding schedule across layers to enable long-range information propagation without repeated or uncoalesced memory access. To further enhance cross-tile communication and positional awareness, HilbertA introduces a small central shared region. Implemented in Triton, HilbertA delivers comparable image quality with significant acceleration over prior methods on Flux.1-dev, demonstrating the feasibility of hardware-aligned two-dimensional sparse attention for high-resolution image generation. HilbertA delivers attention speedups of $2.3\times$ when generating $1024\times 1024$ images, and up to $4.17\times$ at $2048\times 2048$, while achieving image quality comparable to or surpassing baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2509_26538
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle HilbertA: Hilbert Attention for Image Generation with Diffusion Models
Zheng, Shaoyi
Lu, Wenbo
Xia, Yuxuan
Liu, Haomin
Wang, Shengjie
Artificial Intelligence
Designing sparse attention for diffusion transformers requires reconciling two-dimensional spatial locality with GPU efficiency, a trade-off that current methods struggle to achieve. Existing approaches enforce two-dimensional spatial locality but often incur uncoalesced memory access. We present HilbertA, a 2D-aware and GPU-efficient sparse attention mechanism. HilbertA reorders image tokens along Hilbert curves to achieve a contiguous memory layout while preserving spatial neighborhoods, and employs a sliding schedule across layers to enable long-range information propagation without repeated or uncoalesced memory access. To further enhance cross-tile communication and positional awareness, HilbertA introduces a small central shared region. Implemented in Triton, HilbertA delivers comparable image quality with significant acceleration over prior methods on Flux.1-dev, demonstrating the feasibility of hardware-aligned two-dimensional sparse attention for high-resolution image generation. HilbertA delivers attention speedups of $2.3\times$ when generating $1024\times 1024$ images, and up to $4.17\times$ at $2048\times 2048$, while achieving image quality comparable to or surpassing baselines.
title HilbertA: Hilbert Attention for Image Generation with Diffusion Models
topic Artificial Intelligence
url https://arxiv.org/abs/2509.26538