arXiv · ML· Zhida Jiang, Zhaolong Xing, Jiawei Lu, Yipei Niu, Qingyuan Sang, Liangxu Zhang, Wenquan Dai, Junhua Shu, Jiaxing Wang, Qiangyu Pei, Qiong Chen, Xinyu Liu, Fangming Liu, Ai Han, Zhen Chen, Ke Zhang·· 24 天前AI 评分38
FlexMARL:用于基于LLM的多智能体强化学习的图驱动Rollout-Training协同设计框架
FlexMARL: Graph-Driven Rollout-Training Co-Design for LLM-Based Multi-Agent Reinforcement Learning
AI 导读
提出FlexMARL框架,解决大规模基于LLM的多智能体强化学习中的基础设施瓶颈。在生产集群测试中,该框架实现最高7.3倍的推理加速,并将硬件利用率提升高达5.6倍。
来源:arXiv · ML · arxiv.org