arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2604.18529cs.PFcs.DC
收藏

HybridGen: 通过CPU-GPU混合计算实现高效的LLM生成推理

HybridGen: Efficient LLM Generative Inference via CPU-GPU Hybrid Computing

University of California, Merced(加州大学默塞德分校);NVIDIA(英伟达)

Mao Lin, Xi Wang, Guilherme Cox, Dong Li, Hyeran Jeon

AI总结:

HybridGen通过CPU-GPU混合计算解决长上下文LLM推理中的内存带宽和负载不平衡问题,采用注意力logit并行、反馈调度器和语义感知的KV缓存映射,提升效率1.41倍至3.2倍。

摘要:

随着现代LLM支持数千到数百万个token,KV缓存增长到数百GB,对内存容量和带宽构成压力。现有解决方案如KV缓存修剪和卸载虽能缓解问题,但仅依赖GPU或CPU进行注意力计算,且受限于有限的CPU本地内存存储KV缓存。我们提出HybridGen,一种高效的混合注意力框架,用于长上下文LLM推理。HybridGen在具备扩展层级内存的系统中实现CPU-GPU协同注意力,解决三个关键挑战:(1) 多维注意力依赖性,(2) 长序列下CPU-GPU负载不平衡,(3) 层级内存的NUMA惩罚。HybridGen通过引入注意力logit并行、反馈驱动调度器和语义感知的KV缓存映射来解决这些问题。在三种LLM模型、十一种不同规模和三种GPU平台上的实验表明,HybridGen在保持高精度的同时,平均比六种最先进的KV缓存管理方法快1.41倍至3.2倍。

英文摘要:

As modern LLMs support thousands to millions of tokens, KV caches grow to hundreds of gigabytes, stressing memory capacity and bandwidth. Existing solutions, such as KV cache pruning and offloading, alleviate these but underutilize hardware by relying solely on either GPU or CPU for attention computing, and considering yet limited CPU local memory for KV cache storage. We propose HybridGen, an efficient hybrid attention framework for long-context LLM inference. HybridGen enables CPU-GPU collaborative attention on systems with expanded tiered memory (e.g., CXL memory), addressing three key challenges: (1) multi-dimensional attention dependencies, (2) intensifying CPU-GPU load imbalance with longer sequences, and (3) NUMA penalty of tiered memories. HybridGen tackles these by introducing attention logit parallelism, a feedback-driven scheduler, and semantic-aware KV cache mapping. Experiments with three LLM models with eleven different sizes on three GPU platforms with a CXL-expanded memory show that HybridGen outperforms six state-of-the-art KV cache management methods by 1.41x--3.2x on average while maintaining superior accuracy.

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。

↑