English(EN) Network Contention State Aware Dynamic Optimizer for Accelerator Collective Communication Libraries

AI训练优化器将NVIDIA GPU上的AllReduce延迟降低了36%

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-07 21:57

研究人员开发了一种新颖的多智能体系统，用于优化大规模分布式AI/ML训练的集体通信库（CCL）。该系统使用长短期记忆（LSTM）网络检测网络争用，并使用具有保守Q学习的双重深度Q网络（Dueling Double DQN）动态调整CCL配置。该优化器在NVIDIA的NCCL上实现，显著降低了延迟，在NVIDIA A100 GPU的Delta集群上运行的AI工作负载上实现了高达65%的加速。 AI

影响优化分布式训练基础设施，可能加速大规模AI模型的开发和部署。

排序理由详细介绍AI/ML训练基础设施新颖优化技术的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.MA (Multiagent) 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 详细介绍AI/ML训练基础设施新颖优化技术的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: infra, paper, model release; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: 2 days old; Coverage has settled into its steady-state source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.MA (Multiagent) TIER_1 English(EN) · MD S Q Zulkar Nine · 2026-11-07 21:57

面向加速器集体通信库的网络争用状态感知动态优化器

Collective Communication Libraries (CCL) are critical for large-scale distributed AI/ML training, yet their configurations are determined by static heuristics that do not account for real-time network conditions. Our empirical study shows that these heuristics perform near-optima…

本文如何被排名

报道来源 [1]

面向加速器集体通信库的网络争用状态感知动态优化器

相关实体

相关话题