SSD：用于高效自回归图像生成的空间-语义头解耦方法
SSD: Spatial-Semantic Head Decoupling for Efficient Autoregressive Image Generation
AI总结：
针对自回归图像生成模型内存与计算成本高的问题，提出基于空间局部性与涌现语义汇聚现象的SSD KV缓存压缩框架，通过解耦注意力头实现5倍内存降低与6.6倍吞吐量提升，支持受限硬件上的高效图像生成。
摘要：
Janus-Pro等自回归图像生成模型可生成高质量图像，但由于视觉token数量庞大，会带来高内存占用与持续增长的计算需求等显著成本。尽管KV缓存压缩在语言建模领域已被广泛研究，但在图像生成领域仍基本未被探索。在本研究中，我们首先发现了一种独特且显著的注意力现象，将其命名为空间局部性与涌现语义汇聚（semantic sink）。为利用这一关键洞察，我们提出了一种全新的KV缓存压缩框架。具体而言，我们通过将注意力头自适应解耦为两种独立类型，对所有视觉token的KV缓存进行压缩：对于空间局部性头，我们的方法维持一个较短的近期token窗口；对于语义汇聚头，该方法会策略性地保留一组紧凑的高关注度token。我们的大量实验表明，所提方法实现了5倍的内存使用量降低，以及整体吞吐量6.6倍的显著提升，且仅带来极小的视觉质量损失，从而可在资源受限的硬件上实现高效的原生自回归图像生成。
英文摘要：
Autoregressive image generation models like Janus-Pro produce high-quality images, but at the significant cost of high memory and ever-growing computational demands due to the large number of visual tokens. While KV cache compression has been extensively studied in language modeling, it still remains largely unexplored for the image generation domain. In this work, we begin by identifying a distinct and prominent attention phenomenon, which we term spatial locality and emergent semantic sink. To leverage this key insight, we introduce a novel KV cache compression framework. Specifically, we compress the KV cache for all visual tokens by adaptively decoupling attention heads into two separate types: for spatial-locality heads, our method maintains a short recent token window; for semantic-sink heads, it strategically preserves a compact set of highly-attended tokens. Our extensive experiments demonstrate that the proposed method achieves a 5$\times$ reduction in memory usage and a notable 6.6$\times$ speedup in overall throughput with only minimal visual quality loss, thereby enabling highly efficient native autoregressive image generation on resource-constrained hardware.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
