paper-with-me

홈 › Papers

HybridEP: Scaling Expert Parallelism to Cross-Datacenter Scenario via Hybrid Expert/Data Transmission

2025-10-22 · Weihao Yang, Hao Huang, Donglei Wu, Ningke Li, Yanqi Pan, Qiyang Zheng, Wen Xia, Shiyi Li, Qiang Wang arxiv

Mixture-of-Experts (MoE) has become a popular architecture for scaling large models. However, the rapidly growing scale outpaces model training on a single DC, driving a shift toward a more flexible, cross-DC training paradigm. Under this, Expert Parallelism (EP) of MoE faces significant scalability issues due to the limited cross-DC bandwidth. Specifically, existing EP optimizations attempt to overlap data communication and computation, which has little benefit in low-bandwidth scenarios due to a much longer data communication time. Therefore, the trends of cross-DC EP scaling is fast becoming a critical roadblock to the continued growth of MoE models. To address this, we propose HybridEP, a modeling-guided framework to optimize EP under constrained bandwidth. Our key idea is to dynamically transform the spatial placement of experts to reduce data communication traffic and frequency, thereby minimizing EP's communication overheads. However, it is non-trivial to find the optimal solution because it complicates the original communication pattern by mixing data and expert communication. We therefore build a stream-based model to determine the optimal transmission ratio. Guided by this, we incorporate two techniques: (1) domain-based partition to construct the mapping between hybrid patterns and specific communication topology at GPU level, and (2) parameter-efficient migration to further refine this topology by reducing expert transmission overhead and enlarging the domain size. Combining all these designs, HybridEP can be considered as a more general EP with better scalability. Experimental results show that HybridEP outperforms existing state-of-the-art MoE training systems by up to 5.6x under constrained bandwidth. We further compare HybridEP and EP on large-scale simulations. HybridEP achieves up to 1.45x speedup with 1k DCs under different bandwidths.

📄 PDF Abstract BibTeX arXiv:2510.19470

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

InfiniteHBD: Building Datacenter-Scale High-Bandwidth Domain for LLM with Optical Circuit Switching Transceivers

2025-02-06 · Chenchen Shou, Guyue Liu, Hao Nie, Huaiyu Meng 외

Scaling Large Language Model (LLM) training relies on multi-dimensional parallelism, where High-Bandwidth Domains (HBDs) are critical for communication-intensive parallelism like Tensor Parallelism (TP) and Expert Parall…

GPULarge Language Model

NEST: Network- and Memory-Aware Device Placement For Distributed Deep Learning

2026-03-06 · Irene Wang, Vishnu Varma Venkata, Arvind Krishnamurthy, Divya Mahajan arxiv

The growing scale of deep learning demands distributed training frameworks that jointly reason about parallelism, memory, and network topology. Prior works often rely on heuristic or topology-agnostic search, handling co…

Searching for Fast Model Families on Datacenter Accelerators

2021-02-10 · CVPR 2021 1 · Sheng Li, Mingxing Tan, Ruoming Pang, Andrew Li 외

Neural Architecture Search (NAS), together with model scaling, has shown remarkable progress in designing high accuracy and fast convolutional architecture families. However, as neither NAS nor model scaling considers su…

modelNeural Architecture Search

Subspace Networks: Scaling Decentralized Training with Communication-Efficient Model Parallelism

2025-06-02 · Sameera Ramasinghe, Thalaiyasingam Ajanthan, Gil Avraham, Yan Zuo 외 arxiv

Scaling models has led to significant advancements in deep learning, but training these models in decentralized settings remains challenging due to communication bottlenecks. While existing compression techniques are eff…

Cross-Platform Scaling of Vision-Language-Action Models from Edge to Cloud GPUs

2025-09-15 · Amir Taherin, Juyi Lin, Arash Akbari, Arman Akbari 외 arxiv

Vision-Language-Action (VLA) models have emerged as powerful generalist policies for robotic control, yet their performance scaling across model architectures and hardware platforms, as well as their associated power bud…