BiSup: Bidirectional Quantization Error Suppression for Large Language Models

TL;DR

BiSup introduces a bidirectional quantization error suppression method for large language models, effectively reducing error accumulation and diffusion during weight-activation quantization, leading to improved performance and efficiency.

Contribution

The paper proposes BiSup, a novel bidirectional error suppression technique that combines parameter-efficient fine-tuning and prompt mixed-precision strategies for better quantization of LLMs.

Findings

Significant perplexity reduction on Llama and Qwen models.

Outperforms state-of-the-art quantization methods.

Enhances practical deployment of low-bit LLMs.

Abstract

As the size and context length of Large Language Models (LLMs) grow, weight-activation quantization has emerged as a crucial technique for efficient deployment of LLMs. Compared to weight-only quantization, weight-activation quantization presents greater challenges due to the presence of outliers in activations. Existing methods have made significant progress by exploring mixed-precision quantization and outlier suppression. However, these methods primarily focus on optimizing the results of single matrix multiplication, neglecting the bidirectional propagation of quantization errors in LLMs. Specifically, errors accumulate vertically within the same token through layers, and diffuse horizontally across different tokens due to self-attention mechanisms. To address this issue, we introduce BiSup, a Bidirectional quantization error Suppression method. By constructing appropriate optimizable parameter spaces, BiSup utilizes a small amount of data for quantization-aware parameter-efficient fine-tuning to suppress the error vertical accumulation. Besides, BiSup employs prompt mixed-precision quantization strategy, which preserves high precision for the key-value cache of system prompts, to mitigate the error horizontal diffusion. Extensive experiments on Llama and Qwen families demonstrate that BiSup can improve performance over two state-of-the-art methods (the average WikiText2 perplexity decreases from 13.26 to 9.41 for Atom and from 14.33 to 7.85 for QuaRot under the W3A3-g128 configuration), further facilitating the practical applications of low-bit weight-activation quantization.