ProMoE: Fast MoE-based LLM Serving using Proactive Caching

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Song, Xiaoniu, Zhong, Zihang, Chen, Rong, Chen, Haibo
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866916926297473024
author Song, Xiaoniu
Zhong, Zihang
Chen, Rong
Chen, Haibo
author_facet Song, Xiaoniu
Zhong, Zihang
Chen, Rong
Chen, Haibo
contents The promising applications of large language models are often limited by the constrained GPU memory capacity available on edge devices. Mixture-of-Experts (MoE) models help address this issue by activating only a subset of the model's parameters during computation. This approach allows the unused parameters to be offloaded to host memory, thereby reducing the overall GPU memory demand. However, existing cache-based offloading solutions handle cache misses reactively, which significantly impacts system performance. In this paper, we introduce ProMoE, a novel proactive caching system that utilizes intermediate results to predict subsequent expert usage. By proactively fetching experts in advance, ProMoE eliminates passive cache misses, removes loading time from the critical path, and reduces the performance overhead associated with offloading. Our evaluations demonstrate that ProMoE achieves an average speedup of 2.20x (up to 3.21x) and 2.07x (up to 5.02x) in the prefill and decode stages, respectively, compared to existing offloading solutions.
format Preprint
id arxiv_https___arxiv_org_abs_2410_22134
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle ProMoE: Fast MoE-based LLM Serving using Proactive Caching
Song, Xiaoniu
Zhong, Zihang
Chen, Rong
Chen, Haibo
Distributed, Parallel, and Cluster Computing
Artificial Intelligence
The promising applications of large language models are often limited by the constrained GPU memory capacity available on edge devices. Mixture-of-Experts (MoE) models help address this issue by activating only a subset of the model's parameters during computation. This approach allows the unused parameters to be offloaded to host memory, thereby reducing the overall GPU memory demand. However, existing cache-based offloading solutions handle cache misses reactively, which significantly impacts system performance. In this paper, we introduce ProMoE, a novel proactive caching system that utilizes intermediate results to predict subsequent expert usage. By proactively fetching experts in advance, ProMoE eliminates passive cache misses, removes loading time from the critical path, and reduces the performance overhead associated with offloading. Our evaluations demonstrate that ProMoE achieves an average speedup of 2.20x (up to 3.21x) and 2.07x (up to 5.02x) in the prefill and decode stages, respectively, compared to existing offloading solutions.
title ProMoE: Fast MoE-based LLM Serving using Proactive Caching
topic Distributed, Parallel, and Cluster Computing
Artificial Intelligence
url https://arxiv.org/abs/2410.22134