Saved in:
Bibliographic Details
Main Authors: Sun, Manxi, Liu, Wei, Luan, Jian, Gao, Pengzhi, Wang, Bin
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2409.12210
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916400091627520
author Sun, Manxi
Liu, Wei
Luan, Jian
Gao, Pengzhi
Wang, Bin
author_facet Sun, Manxi
Liu, Wei
Luan, Jian
Gao, Pengzhi
Wang, Bin
contents The Sparsely-Activated Mixture-of-Experts (MoE) has gained increasing popularity for scaling up large language models (LLMs) without exploding computational costs. Despite its success, the current design faces a challenge where all experts have the same size, limiting the ability of tokens to choose the experts with the most appropriate size for generating the next token. In this paper, we propose the Mixture of Diverse Size Experts (MoDSE), a new MoE architecture with layers designed to have experts of different sizes. Our analysis of difficult token generation tasks shows that experts of various sizes achieve better predictions, and the routing path of the experts tends to be stable after a training period. However, having experts of diverse sizes can lead to uneven workload distribution. To tackle this limitation, we introduce an expert-pair allocation strategy to evenly distribute the workload across multiple GPUs. Comprehensive evaluations across multiple benchmarks demonstrate the effectiveness of MoDSE, as it outperforms existing MoEs by allocating the parameter budget to experts adaptively while maintaining the same total parameter size and the number of experts.
format Preprint
id arxiv_https___arxiv_org_abs_2409_12210
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Mixture of Diverse Size Experts
Sun, Manxi
Liu, Wei
Luan, Jian
Gao, Pengzhi
Wang, Bin
Machine Learning
Artificial Intelligence
The Sparsely-Activated Mixture-of-Experts (MoE) has gained increasing popularity for scaling up large language models (LLMs) without exploding computational costs. Despite its success, the current design faces a challenge where all experts have the same size, limiting the ability of tokens to choose the experts with the most appropriate size for generating the next token. In this paper, we propose the Mixture of Diverse Size Experts (MoDSE), a new MoE architecture with layers designed to have experts of different sizes. Our analysis of difficult token generation tasks shows that experts of various sizes achieve better predictions, and the routing path of the experts tends to be stable after a training period. However, having experts of diverse sizes can lead to uneven workload distribution. To tackle this limitation, we introduce an expert-pair allocation strategy to evenly distribute the workload across multiple GPUs. Comprehensive evaluations across multiple benchmarks demonstrate the effectiveness of MoDSE, as it outperforms existing MoEs by allocating the parameter budget to experts adaptively while maintaining the same total parameter size and the number of experts.
title Mixture of Diverse Size Experts
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2409.12210