Rethinking 1-bit Optimization Leveraging Pre-trained Large Language Models
This enables efficient deployment of large language models with reduced storage and computational costs while maintaining performance, though it's an incremental improvement over existing quantization methods.
The paper tackles the problem of 1-bit LLM quantization causing accuracy degradation when training from scratch, and introduces a progressive training method with binary-aware initialization and dual-scaling compensation that outperforms existing approaches on various LLM sizes.
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 direct adaptation difficult. In this paper, we introduce a consistent progressive training for both forward and backward, smoothly converting the floating-point 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.