ACE: Exploring Activation Cosine Similarity and Variance for Accurate and Calibration-Efficient LLM Pruning

TL;DR

This paper introduces a novel LLM pruning method that combines activation cosine similarity and variance metrics to improve accuracy, calibration, and efficiency, achieving significant reductions in perplexity and pruning time.

Contribution

It proposes a new pruning approach using activation cosine similarity and variance metrics, enhancing both performance and speed of LLM pruning.

Findings

Achieves up to 18% perplexity reduction.

Reduces pruning time by up to 63%.

Effective on models like LLaMA, LLaMA-2, and OPT.

Abstract

With the rapid expansion of large language models (LLMs), the demand for memory and computational resources has grown significantly. Recent advances in LLM pruning aim to reduce the size and computational cost of these models. However, existing methods often suffer from either suboptimal pruning performance or low time efficiency during the pruning process. In this work, we propose an efficient and effective pruning method that simultaneously achieves high pruning performance and fast pruning speed with improved calibration efficiency. Our approach introduces two key innovations: (1) An activation cosine similarity loss-guided pruning metric, which considers the angular deviation of the output activation between the dense and pruned models. (2) An activation variance-guided pruning metric, which helps preserve semantic distinctions in output activations after pruning, enabling effective pruning with shorter input sequences. These two components can be readily combined to enhance LLM pruning in both accuracy and efficiency. Experimental results show that our method achieves up to an 18% reduction in perplexity and up to 63% decrease in pruning time on prevalent LLMs such as LLaMA, LLaMA-2, and OPT.