GPTailor: Large Language Model Pruning Through Layer Cutting and Stitching

TL;DR

This paper introduces GPTailor, a novel method for compressing large language models by layer cutting and stitching, which maintains high performance while significantly reducing model size.

Contribution

The paper presents a new approach to LLM pruning that combines layer removal, selection, and merging from finetuned variants, optimizing model compression.

Findings

Maintains 97.3% of original performance on Llama2-13B after 25% parameter reduction.

Outperforms previous state-of-the-art pruning methods.

Uses a zero-order optimization framework for layer operations.

Abstract

Large language models (LLMs) have shown remarkable capabilities in language understanding and generation. However, such impressive capability typically comes with a substantial model size, which presents significant challenges in deployment and inference. While structured pruning of model parameters offers a promising way to reduce computational costs at deployment time, current methods primarily focus on single model pruning. In this work, we develop a novel strategy to compress models by strategically combining or merging layers from finetuned model variants, which preserves the original model's abilities by aggregating capabilities accentuated in different finetunes. We pose the optimal tailoring of these LLMs as a zero-order optimization problem, adopting a search space that supports three different operations: (1) Layer removal, (2) Layer selection from different candidate models, and (3) Layer merging. Our experiments demonstrate that this approach leads to competitive model pruning, for example, for the Llama2-13B model families, our compressed models maintain approximately 97.3\% of the original performance while removing $\sim25\%$ of parameters, significantly outperforming previous state-of-the-art methods. The code is available at https://github.com/Guinan-Su/auto-merge-llm.