LLaMA-NAS: Efficient Neural Architecture Search for Large Language Models

TL;DR

This paper introduces LLaMA-NAS, a one-shot neural architecture search method that efficiently finds smaller, high-performing LLM architectures based on LLaMA2-7B, reducing size and computational costs with minimal accuracy loss.

Contribution

It presents a novel NAS approach using genetic algorithms to optimize LLaMA2-7B architectures, achieving significant size and speed improvements over existing pruning methods.

Findings

1.5x reduction in model size

1.3x increase in throughput

Comparable accuracy with smaller models

Abstract

The abilities of modern large language models (LLMs) in solving natural language processing, complex reasoning, sentiment analysis and other tasks have been extraordinary which has prompted their extensive adoption. Unfortunately, these abilities come with very high memory and computational costs which precludes the use of LLMs on most hardware platforms. To mitigate this, we propose an effective method of finding Pareto-optimal network architectures based on LLaMA2-7B using one-shot NAS. In particular, we fine-tune LLaMA2-7B only once and then apply genetic algorithm-based search to find smaller, less computationally complex network architectures. We show that, for certain standard benchmark tasks, the pre-trained LLaMA2-7B network is unnecessarily large and complex. More specifically, we demonstrate a 1.5x reduction in model size and 1.3x speedup in throughput for certain tasks with negligible drop in accuracy. In addition to finding smaller, higher-performing network architectures, our method does so more effectively and efficiently than certain pruning or sparsification techniques. Finally, we demonstrate how quantization is complementary to our method and that the size and complexity of the networks we find can be further decreased using quantization. We believe that our work provides a way to automatically create LLMs which can be used on less expensive and more readily available hardware platforms.