DReSS: Data-driven Regularized Structured Streamlining for Large Language Models
This addresses the challenge of efficiently compressing LLMs for deployment, offering a novel paradigm that mitigates performance degradation in pruning, though it is incremental in the context of model compression techniques.
The paper tackles the problem of high computational and memory costs in large language models by proposing DReSS, a method that applies regularization before pruning to reduce information loss, resulting in significant performance improvements over existing pruning methods even under extreme pruning ratios, with reduced latency and increased throughput.
Large language models (LLMs) have achieved significant progress across various domains, but their increasing scale results in high computational and memory costs. Recent studies have revealed that LLMs exhibit sparsity, providing the potential to reduce model size through pruning techniques. However, existing pruning methods typically follow a prune-then-finetune paradigm. Since the pruned components still contain valuable information, their direct removal often leads to irreversible performance degradation, imposing a substantial computational burden to recover performance during finetuning. In this paper, we propose a novel paradigm that first applies regularization, then prunes, and finally finetunes. Based on this paradigm, we introduce DReSS, a simple and effective Data-driven Regularized Structured Streamlining method for LLMs. By leveraging a small amount of data to regularize the components to be pruned, DReSS explicitly transfers the important information to the remaining parts of the model in advance. Compared to direct pruning, this can reduce the information loss caused by parameter removal, thereby enhancing its language modeling capabilities. Experimental results demonstrate that DReSS significantly outperforms existing pruning methods even under extreme pruning ratios, significantly reducing latency and increasing throughput.