ExpertWeave: Efficiently Serving Expert-Specialized Fine-Tuned Adapters at Scale

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Shi, Ge, Sadri, Hanieh, Wang, Qian, Zhang, Yu, Xiong, Ying, Zhang, Yong, Fan, Zhenan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912552040005632
author Shi, Ge
Sadri, Hanieh
Wang, Qian
Zhang, Yu
Xiong, Ying
Zhang, Yong
Fan, Zhenan
author_facet Shi, Ge
Sadri, Hanieh
Wang, Qian
Zhang, Yu
Xiong, Ying
Zhang, Yong
Fan, Zhenan
contents Expert-Specialized Fine-Tuning (ESFT) adapts Mixture-of-Experts (MoE) large language models to enhance their task-specific performance by selectively tuning the top-activated experts for the task. Serving these fine-tuned models at scale is challenging: deploying merged models in isolation is prohibitively resource-hungry, while existing multi-adapter serving systems with LoRA-style additive updates are incompatible with ESFT's expert-oriented paradigm. We present ExpertWeave, a system that serves multiple ESFT adapters concurrently over a single shared MoE base model, drastically reducing the memory footprint and improving resource utilization. To seamlessly integrate into existing inference pipelines for MoE models with non-intrusive modifications and minimal latency overhead, ExpertWeave introduces a virtual-memory-assisted expert weight manager that co-locates base-model and adapter experts without incurring memory overhead from fragmentation, and a fused kernel for batched rerouting to enable lightweight redirection of tokens to the appropriate experts at runtime. Our evaluations show that ExpertWeave can simultaneously serve multiple adapters of a 16B MoE model on a single accelerator where the baseline runs out of memory, or provides up to 94x more KV cache capacity and achieves up to 18% higher throughput while using comparable resources, all without compromising model accuracy. ExpertWeave maintains low overhead even when scaling to 20 adapters, with a 4-11% latency increase compared with serving the base model alone. Source code will be released soon.
format Preprint
id arxiv_https___arxiv_org_abs_2508_17624
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ExpertWeave: Efficiently Serving Expert-Specialized Fine-Tuned Adapters at Scale
Shi, Ge
Sadri, Hanieh
Wang, Qian
Zhang, Yu
Xiong, Ying
Zhang, Yong
Fan, Zhenan
Distributed, Parallel, and Cluster Computing
Expert-Specialized Fine-Tuning (ESFT) adapts Mixture-of-Experts (MoE) large language models to enhance their task-specific performance by selectively tuning the top-activated experts for the task. Serving these fine-tuned models at scale is challenging: deploying merged models in isolation is prohibitively resource-hungry, while existing multi-adapter serving systems with LoRA-style additive updates are incompatible with ESFT's expert-oriented paradigm. We present ExpertWeave, a system that serves multiple ESFT adapters concurrently over a single shared MoE base model, drastically reducing the memory footprint and improving resource utilization. To seamlessly integrate into existing inference pipelines for MoE models with non-intrusive modifications and minimal latency overhead, ExpertWeave introduces a virtual-memory-assisted expert weight manager that co-locates base-model and adapter experts without incurring memory overhead from fragmentation, and a fused kernel for batched rerouting to enable lightweight redirection of tokens to the appropriate experts at runtime. Our evaluations show that ExpertWeave can simultaneously serve multiple adapters of a 16B MoE model on a single accelerator where the baseline runs out of memory, or provides up to 94x more KV cache capacity and achieves up to 18% higher throughput while using comparable resources, all without compromising model accuracy. ExpertWeave maintains low overhead even when scaling to 20 adapters, with a 4-11% latency increase compared with serving the base model alone. Source code will be released soon.
title ExpertWeave: Efficiently Serving Expert-Specialized Fine-Tuned Adapters at Scale
topic Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2508.17624