Enregistré dans:
Détails bibliographiques
Auteurs principaux: Bu, Tao, Wang, Qiangang, Zeng, Bowen, Sun, Hanwen, Huang, Yunpeng, Cao, Chun, Xu, Jingwei
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2510.17896
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866915565504823296
author Bu, Tao
Wang, Qiangang
Zeng, Bowen
Sun, Hanwen
Huang, Yunpeng
Cao, Chun
Xu, Jingwei
author_facet Bu, Tao
Wang, Qiangang
Zeng, Bowen
Sun, Hanwen
Huang, Yunpeng
Cao, Chun
Xu, Jingwei
contents Transformer-based large language models (LLMs) have achieved remarkable success, yet their standard attention mechanism incurs quadratic computation and memory costs with respect to sequence length, posing a major bottleneck for long-context training. Prior work tackles this challenge along two directions: (1) kernel-level optimizations, which accelerate dense and sparse attention operators; and (2) module-level strategies, often referred to as distributed attention or context parallel training, which scale attention across multiple devices. However, systematic evaluation still remains limited: operator-level comparisons are often incomplete, while context parallel strategies are typically framework-specific, with unclear performance analysis across contexts. To address these gaps, we propose a unified benchmark that integrates representative attention kernels and context parallel mechanisms with a modular and extensible interface for evaluation. The benchmark evaluates methods along two critical dimensions: (1) attention mask patterns, which strongly affect efficiency, scalability, and usability, and (2) sequence length and distributed scale, which determine performance under extreme long-context training. Through comprehensive experiments on the cluster of up to 96 GPUs, our benchmark enables reproducible comparisons, highlights method-specific trade-offs, and provides practical guidance for designing and deploying attention mechanisms in long-context LLM training.
format Preprint
id arxiv_https___arxiv_org_abs_2510_17896
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Long-Context Attention Benchmark: From Kernel Efficiency to Distributed Context Parallelism
Bu, Tao
Wang, Qiangang
Zeng, Bowen
Sun, Hanwen
Huang, Yunpeng
Cao, Chun
Xu, Jingwei
Machine Learning
Artificial Intelligence
Transformer-based large language models (LLMs) have achieved remarkable success, yet their standard attention mechanism incurs quadratic computation and memory costs with respect to sequence length, posing a major bottleneck for long-context training. Prior work tackles this challenge along two directions: (1) kernel-level optimizations, which accelerate dense and sparse attention operators; and (2) module-level strategies, often referred to as distributed attention or context parallel training, which scale attention across multiple devices. However, systematic evaluation still remains limited: operator-level comparisons are often incomplete, while context parallel strategies are typically framework-specific, with unclear performance analysis across contexts. To address these gaps, we propose a unified benchmark that integrates representative attention kernels and context parallel mechanisms with a modular and extensible interface for evaluation. The benchmark evaluates methods along two critical dimensions: (1) attention mask patterns, which strongly affect efficiency, scalability, and usability, and (2) sequence length and distributed scale, which determine performance under extreme long-context training. Through comprehensive experiments on the cluster of up to 96 GPUs, our benchmark enables reproducible comparisons, highlights method-specific trade-offs, and provides practical guidance for designing and deploying attention mechanisms in long-context LLM training.
title Long-Context Attention Benchmark: From Kernel Efficiency to Distributed Context Parallelism
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2510.17896