Greedy Output Approximation: Towards Efficient Structured Pruning for LLMs Without Retraining

TL;DR

This paper introduces a novel single-shot pruning method for large language models that avoids retraining, using a simplified structure and inference-aware criteria to efficiently reduce model size while preserving performance.

Contribution

It proposes a depth-2 pruning structure and inference-aware criteria for efficient, retraining-free structured pruning of LLMs, improving over traditional metrics.

Findings

Significantly reduces computational costs and hardware needs.

Maintains high performance across multiple datasets and models.

Outperforms traditional training-aware pruning metrics.

Abstract

To remove redundant components of large language models (LLMs) without incurring significant computational costs, this work focuses on single-shot pruning without a retraining phase. We simplify the pruning process for Transformer-based LLMs by identifying a depth-2 pruning structure that functions independently. Additionally, we propose two inference-aware pruning criteria derived from the optimization perspective of output approximation, which outperforms traditional training-aware metrics such as gradient and Hessian. We also introduce a two-step reconstruction technique to mitigate pruning errors without model retraining. Experimental results demonstrate that our approach significantly reduces computational costs and hardware requirements while maintaining superior performance across various datasets and models.