Papers
arxiv:2505.17639

PreMoe: Lightening MoEs on Constrained Memory by Expert Pruning and Retrieval

Published on May 23
· Submitted by Eleven-P on May 28
Authors:
,
,
,
,
,
,

Abstract

PreMoe framework enables efficient deployment of large MoE language models in memory-constrained environments by pruning and retrieving task-specific experts.

AI-generated summary

Mixture-of-experts (MoE) architectures enable scaling large language models (LLMs) to vast parameter counts without a proportional rise in computational costs. However, the significant memory demands of large MoE models hinder their deployment across various computational environments, from cloud servers to consumer devices. This study first demonstrates pronounced task-specific specialization in expert activation patterns within MoE layers. Building on this, we introduce PreMoe, a novel framework that enables efficient deployment of massive MoE models in memory-constrained environments. PreMoe features two main components: probabilistic expert pruning (PEP) and task-adaptive expert retrieval (TAER). PEP employs a new metric, the task-conditioned expected selection score (TCESS), derived from router logits to quantify expert importance for specific tasks, thereby identifying a minimal set of critical experts. TAER leverages these task-specific expert importance profiles for efficient inference. It pre-computes and stores compact expert patterns for diverse tasks. When a user query is received, TAER rapidly identifies the most relevant stored task pattern and reconstructs the model by loading only the small subset of experts crucial for that task. This approach dramatically reduces the memory footprint across all deployment scenarios. DeepSeek-R1 671B maintains 97.2\% accuracy on MATH500 when pruned to 8/128 configuration (50\% expert reduction), and still achieves 72.0\% with aggressive 8/32 pruning (87.5\% expert reduction). Pangu-Ultra-MoE 718B achieves 97.15\% on MATH500 and 81.3\% on AIME24 with 8/128 pruning, while even more aggressive pruning to 4/64 (390GB memory) preserves 96.95\% accuracy on MATH500. We make our code publicly available at https://github.com/JarvisPei/PreMoe.

Community

Paper author Paper submitter

The code is available at: https://github.com/JarvisPei/PreMoe

Sign up or log in to comment

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2505.17639 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2505.17639 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2505.17639 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.