AlphaLoRA: Assigning LoRA Experts Based on Layer Training Quality
This addresses redundancy in MoE-LoRA architectures for LLM fine-tuning, offering an incremental improvement in efficiency and performance.
The paper tackles redundancy in combining LoRA with Mixture-of-Experts for parameter-efficient fine-tuning in LLMs by introducing AlphaLoRA, a training-free method that allocates LoRA experts based on layer training quality, achieving comparable or superior performance on ten benchmarks.
Parameter-efficient fine-tuning methods, such as Low-Rank Adaptation (LoRA), are known to enhance training efficiency in Large Language Models (LLMs). Due to the limited parameters of LoRA, recent studies seek to combine LoRA with Mixture-of-Experts (MoE) to boost performance across various tasks. However, inspired by the observed redundancy in traditional MoE structures, previous studies identify similar redundancy among LoRA experts within the MoE architecture, highlighting the necessity for non-uniform allocation of LoRA experts across different layers. In this paper, we leverage Heavy-Tailed Self-Regularization (HT-SR) Theory to design a fine-grained allocation strategy. Our analysis reveals that the number of experts per layer correlates with layer training quality, which exhibits significant variability across layers. Based on this, we introduce AlphaLoRA, a theoretically principled and training-free method for allocating LoRA experts to further mitigate redundancy. Experiments on three models across ten language processing and reasoning benchmarks demonstrate that AlphaLoRA achieves comparable or superior performance over all baselines. Our code is available at https://github.com/morelife2017/alphalora.