paper-with-me

Papers

MegaScale-MoE: Large-Scale Communication-Efficient Training of Mixture-of-Experts Models in Production

2025-05-16 · Chao Jin, Ziheng Jiang, Zhihao Bai, Zheng Zhong, Juncai Liu, Xiang Li, Ningxin Zheng, Xi Wang, Cong Xie, Qi Huang, Wen Heng, Yiyuan Ma, Wenlei Bao, Size Zheng, Yanghua Peng, Haibin Lin, Xuanzhe Liu, Xin Jin, Xin Liu

We present MegaScale-MoE, a production system tailored for the efficient training of large-scale mixture-of-experts (MoE) models. MoE emerges as a promising architecture to scale large language models (LLMs) to unprecedented sizes, thereby enhancing model performance. However, existing MoE training systems experience a degradation in training efficiency, exacerbated by the escalating scale of MoE models and the continuous evolution of hardware. Recognizing the pivotal role of efficient communication in enhancing MoE training, MegaScale-MoE customizes communication-efficient parallelism strategies for attention and FFNs in each MoE layer and adopts a holistic approach to overlap communication with computation at both inter- and intra-operator levels. Additionally, MegaScale-MoE applies communication compression with adjusted communication patterns to lower precision, further improving training efficiency. When training a 352B MoE model on 1,440 NVIDIA Hopper GPUs, MegaScale-MoE achieves a training throughput of 1.41M tokens/s, improving the efficiency by 1.88$\times$ compared to Megatron-LM. We share our operational experience in accelerating MoE training and hope that by offering our insights in system design, this work will motivate future research in MoE systems.

📄 PDF Abstract BibTeX arXiv:2505.11432

Code (0)

등록된 구현이 없습니다.

Tasks

Mixture-of-Experts

Methods 이 논문이 사용한 방법론

Softmax The Softmax output function transforms a previous layer's output into a vector of probabilities. It is commonly used for multiclass classification. Given an input vector $x$…
Attention 설명 없음
MoE 설명 없음

Similar Papers 제목 키워드 기반

MegaScale-Infer: Serving Mixture-of-Experts at Scale with Disaggregated Expert Parallelism

2025-04-03 · Ruidong Zhu, Ziheng Jiang, Chao Jin, Peng Wu 외

Mixture-of-Experts (MoE) showcases tremendous potential to scale large language models (LLMs) with enhanced performance and reduced computational complexity. However, its sparsely activated architecture shifts feed-forwa…

CPUGPUMixture-of-Experts

MegaScale: Scaling Large Language Model Training to More Than 10,000 GPUs

2024-02-23 · Ziheng Jiang, Haibin Lin, Yinmin Zhong, Qi Huang 외

We present the design, implementation and engineering experience in building and deploying MegaScale, a production system for training large language models (LLMs) at the scale of more than 10,000 GPUs. Training LLMs at …

Language ModelingLanguage ModellingLarge Language Model

Making MoE-based LLM Inference Resilient with Tarragon

2026-01-04 · Songyu Zhang, Aaron Tam, Myungjin Lee, Shixiong Qi 외 arxiv

Mixture-of-Experts (MoE) models are increasingly used to serve LLMs at scale, but failures become common as deployment scale grows. Existing systems exhibit poor failure resilience: even a single worker failure triggers …

Efficient Training of Large-Scale AI Models Through Federated Mixture-of-Experts: A System-Level Approach

2025-07-08 · Xiaobing Chen, Boyang Zhang, Xiangwei Zhou, Mingxuan Sun 외

The integration of Federated Learning (FL) and Mixture-of-Experts (MoE) presents a compelling pathway for training more powerful, large-scale artificial intelligence models (LAMs) on decentralized data while preserving p…

Edge-computingFederated LearningMixture-of-Experts

Mixture-of-Parallelisms: Towards Memory-Efficient Training Stack for Mixture-of-Experts Models

2026-07-02 · Xuan-Phi Nguyen, Shrey Pandit, Yiran Zhao, Semih Yavuz 외 arxiv

This paper showcases a memory-efficient training stack for Mixture-of-Experts (MoE) models. It is a training paradigm that combines and specializes various existing and novel parallelism techniques at different layers an…