A Stronger Mixture of Low-Rank Experts for Fine-Tuning Foundation Models

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Sun, Mengyang, Wang, Yihao, Feng, Tao, Zhang, Dan, Zhu, Yifan, Tang, Jie
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912241480105984
author Sun, Mengyang
Wang, Yihao
Feng, Tao
Zhang, Dan
Zhu, Yifan
Tang, Jie
author_facet Sun, Mengyang
Wang, Yihao
Feng, Tao
Zhang, Dan
Zhu, Yifan
Tang, Jie
contents In order to streamline the fine-tuning of foundation models, Low-Rank Adapters (LoRAs) have been substantially adopted across various fields, including instruction tuning and domain adaptation. The underlying concept of LoRA involves decomposing a full-rank matrix into the product of two lower-rank matrices, which reduces storage consumption and accelerates the training process. Furthermore, to address the limited expressive capacity of LoRA, the Mixture-of-Expert (MoE) has been introduced for incorporating multiple LoRA adapters. The integration of LoRA experts leads to a visible improvement across several downstream scenes. However, the mixture of LoRAs (MoE-LoRA) still exhibits its low robustness during tuning and inferring. Inspired by the Riemannian Preconditioners which train LoRA as a sub-space projector, we propose a new training strategy for MoE-LoRA, to stabilize and boost its feature learning procedure by multi-space projections. Examinations on SGD and AdamW optimizers demonstrate the effectiveness of our methodology. Source code is available at https://github.com/THUDM/MoELoRA_Riemannian.
format Preprint
id arxiv_https___arxiv_org_abs_2502_15828
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Stronger Mixture of Low-Rank Experts for Fine-Tuning Foundation Models
Sun, Mengyang
Wang, Yihao
Feng, Tao
Zhang, Dan
Zhu, Yifan
Tang, Jie
Machine Learning
Artificial Intelligence
In order to streamline the fine-tuning of foundation models, Low-Rank Adapters (LoRAs) have been substantially adopted across various fields, including instruction tuning and domain adaptation. The underlying concept of LoRA involves decomposing a full-rank matrix into the product of two lower-rank matrices, which reduces storage consumption and accelerates the training process. Furthermore, to address the limited expressive capacity of LoRA, the Mixture-of-Expert (MoE) has been introduced for incorporating multiple LoRA adapters. The integration of LoRA experts leads to a visible improvement across several downstream scenes. However, the mixture of LoRAs (MoE-LoRA) still exhibits its low robustness during tuning and inferring. Inspired by the Riemannian Preconditioners which train LoRA as a sub-space projector, we propose a new training strategy for MoE-LoRA, to stabilize and boost its feature learning procedure by multi-space projections. Examinations on SGD and AdamW optimizers demonstrate the effectiveness of our methodology. Source code is available at https://github.com/THUDM/MoELoRA_Riemannian.
title A Stronger Mixture of Low-Rank Experts for Fine-Tuning Foundation Models
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2502.15828