Towards Federated Low-Rank Adaptation of Language Models with Rank Heterogeneity
This work addresses performance instability in federated learning for language models, which is an incremental improvement over existing methods.
The paper tackled the problem of unstable performance in federated low-rank adaptation of language models due to rank heterogeneity among clients, and the result was a replication-based padding strategy that accelerated convergence and enhanced predictive performance.
Low-rank adaptation (LoRA) offers an efficient alternative to full-weight adaptation in federated fine-tuning of language models, significantly reducing computational costs. By adjusting ranks for each client, federated LoRA enables flexible resource allocation. However, we observe that heterogeneous ranks among clients lead to unstable performance. Our analysis attributes this instability to the conventional zero-padding aggregation strategy, which dilutes information from high-rank clients during model aggregation. To address this issue, we propose a replication-based padding strategy that better retains valuable information from clients with high-quality data. Empirically, this approach accelerates convergence and enhances the global model's predictive performance.