AFA-LoRA: Enabling Non-Linear Adaptations in LoRA with Activation Function Annealing
This work addresses a bottleneck in parameter-efficient fine-tuning for machine learning practitioners by enabling more powerful adaptations while maintaining mergeability.
The paper tackled the limited expressive power of linear adaptation in LoRA by proposing AFA-LoRA, a training strategy that introduces non-linear expressivity through an annealed activation function, reducing the performance gap between LoRA and full-parameter training in tasks like supervised fine-tuning and reinforcement learning.
Low-Rank Adaptation (LoRA) is a widely adopted parameter-efficient fine-tuning (PEFT) method. However, its linear adaptation process limits its expressive power. This means there is a gap between the expressive power of linear training and non-linear training. To bridge this gap, we propose AFA-LoRA, a novel training strategy that brings non-linear expressivity to LoRA while maintaining its seamless mergeability. Our key innovation is an annealed activation function that transitions from a non-linear to a linear transformation during training, allowing the adapter to initially adopt stronger representational capabilities before converging to a mergeable linear form. We implement our method on supervised fine-tuning, reinforcement learning, and speculative decoding. The results show that AFA-LoRA reduces the performance gap between LoRA and full-parameter training. This work enables a more powerful and practical paradigm of parameter-efficient adaptation.