Rethinking 1-bit Optimization Leveraging Pre-trained Large Language Models

TL;DR

This paper presents a novel progressive training method for 1-bit quantization of large language models, effectively leveraging pre-trained models to reduce costs and improve accuracy.

Contribution

It introduces a consistent progressive training approach with binary-aware initialization and dual-scaling, enabling high-performance 1-bit LLMs from pre-trained models.

Findings

Outperforms existing 1-bit LLM quantization methods.

Reduces training costs by leveraging pre-trained models.

Achieves high accuracy with 1-bit quantized LLMs.

Abstract

1-bit LLM quantization offers significant advantages in reducing storage and computational costs. However, existing methods typically train 1-bit LLMs from scratch, failing to fully leverage pre-trained models. This results in high training costs and notable accuracy degradation. We identify that the large gap between full precision and 1-bit representations makes naive adaptation difficult. In this paper, we introduce a consistent progressive training for both forward and backward, smoothly converting the full-precision weights into the binarized ones. Additionally, we incorporate binary-aware initialization and dual-scaling compensation to reduce the difficulty of progressive training and improve the performance. Experimental results on LLMs of various sizes demonstrate that our method outperforms existing approaches. Our results show that high-performance 1-bit LLMs can be achieved using pre-trained models, eliminating the need for expensive training from scratch.