CUCo自动化协同优化大模型训练推理的计算与通信，提升效率并降低成本。
CUCo: An Agentic Framework for Compute and Communication Co-design
- 用智能体框架自动设计计算通信协同策略，分快慢路径保证可靠性与性能。
- 在四个多GPU任务中最高提速1.57倍，发现隐藏调度开销的新重叠机制。
- 适合追求高效分布式大模型部署的研发团队，尤其关注算力与通信协同优化者。
分布式大模型训练与推理中的计算与通信传统上独立优化；尽管DeepEP、FLUX、TokenWeave等专家系统展示了协同设计潜力，但仍需深厚系统知识和硬件调优；本文提出CUCo——一个智能体框架，通过结构化设计空间建模，结合以正确性为先的快速路径代理生成可靠基线，以及驱动进化的慢速路径代理探索高性能策略，实现跨四个多GPU工作负载最高1.57倍加速，并在DeepSeek-V3 MoE层发现一种双流重叠策略，将调度开销隐藏于本地计算之中，单次工作负载推理成本低于10美元。
原文摘要 · Abstract (English)
Computation and communication in distributed LLM training and inference are traditionally optimized in isolation; expert-crafted systems such as DeepEP, FLUX, and TokenWeave show the potential of co-design but require deep systems expertise and hardware-specific tuning; CUCo is an agentic framework that automates compute-communication co-design of CUDA kernels by combining a structured design-space formalization with a correctness-first fast-path agent for reliable baselines and an evolution-driven slow-path agent for high-performance strategies, achieving up to 1.57x speedup across four multi-GPU workloads and discovering a two-stream overlap strategy on a DeepSeek-V3 MoE layer that hides dispatch behind local compute at an LLM inference cost under $10 per workload.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。