Dynamic Mixture of Experts: An Auto-Tuning Approach for Efficient Transformer Models

2024-05-23

arXiv.org

AI Summary

Sparse Mixture-of-Experts (MoE) models critically depend on hyperparameters such as the total number of experts and the top-k activation count, making manual tuning computationally expensive. To address this, we propose Dynamic Mixture-of-Experts (DynMoE), the first framework enabling token-level adaptive top-k gating and dynamic expert addition/removal during training—eliminating reliance on fixed hyperparameters. Methodologically, DynMoE introduces a gradient-aware dynamic gating network and an expert pruning mechanism, jointly optimizing sparse activation, load balancing, and parameter efficiency. Evaluated across vision, language, and vision-language multimodal tasks, DynMoE matches the performance of GMoE and MoE-LLaVA while significantly reducing activated parameter count and improving both training and inference efficiency.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Computer Vision: Large Vision Models
• Natural Language Processing: (Large) Language Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

The Sparse Mixture of Experts (SMoE) has been widely employed to enhance the efficiency of training and inference for Transformer-based foundational models, yielding promising results.However, the performance of SMoE heavily depends on the choice of hyper-parameters, such as the number of experts and the number of experts to be activated (referred to as top-k), resulting in significant computational overhead due to the extensive model training by searching over various hyper-parameter configurations. As a remedy, we introduce the Dynamic Mixture of Experts (DynMoE) technique. DynMoE incorporates (1) a novel gating method that enables each token to automatically determine the number of experts to activate. (2) An adaptive process automatically adjusts the number of experts during training. Extensive numerical results across Vision, Language, and Vision-Language tasks demonstrate the effectiveness of our approach to achieve competitive performance compared to GMoE for vision and language tasks, and MoE-LLaVA for vision-language tasks, while maintaining efficiency by activating fewer parameters. Our code is available at https://github.com/LINs-lab/DynMoE.

Problem

• Optimizes hyper-parameters in Sparse Mixture of Experts.
• Reduces computational overhead in Transformer models.
• Enhances efficiency by dynamically adjusting expert activation.

Innovation

• Dynamic Mixture of Experts (DynMoE) technique
• Novel gating method for expert activation
• Adaptive process for expert number adjustment