A Simple and Effective Pruning Approach for Large Language Models

TL;DR

Wanda is a simple, retraining-free pruning method for large language models that prunes weights based on their magnitude and input activations, effectively inducing sparsity without performance loss.

Contribution

Introduces Wanda, a novel pruning approach that does not require retraining or weight updates, leveraging input activations to prune weights in pretrained LLMs.

Findings

Wanda outperforms magnitude pruning baselines on LLaMA models.

Wanda performs competitively with methods requiring weight updates.

The method is effective across various language benchmarks.

Abstract

As their size increases, Large Languages Models (LLMs) are natural candidates for network pruning methods: approaches that drop a subset of network weights while striving to preserve performance. Existing methods, however, require either retraining, which is rarely affordable for billion-scale LLMs, or solving a weight reconstruction problem reliant on second-order information, which may also be computationally expensive. In this paper, we introduce a novel, straightforward yet effective pruning method, termed Wanda (Pruning by Weights and activations), designed to induce sparsity in pretrained LLMs. Motivated by the recent observation of emergent large magnitude features in LLMs, our approach prunes weights with the smallest magnitudes multiplied by the corresponding input activations, on a per-output basis. Notably, Wanda requires no retraining or weight update, and the pruned LLM can be used as is. We conduct a thorough evaluation of our method Wanda on LLaMA and LLaMA-2 across various language benchmarks. Wanda significantly outperforms the established baseline of magnitude pruning and performs competitively against recent method involving intensive weight update. Code is available at https://github.com/locuslab/wanda.