K-Quantization and its Impact on Output Performance

TL;DR

This paper examines how different levels of quantization from 2 to 8 bits affect the performance of large language models across various NLP tasks, highlighting the trade-offs between efficiency and accuracy.

Contribution

It provides a comprehensive analysis of quantization effects on multiple LLMs, revealing how model size and task type influence performance degradation at lower precisions.

Findings

Higher precision (e.g., 8-bit) improves performance with diminishing returns.

Aggressive quantization (e.g., 2-bit) often retains acceptable accuracy for some models.

Larger models are more resilient to lower-bit quantization, especially in mid-sized ranges.

Abstract

Recent advancements in large language models (LLMs) have shown their remarkable capacities in many NLP tasks. However, their substantial size often presents challenges for deployment. This necessitates efficient techniques for model compression, with quantization emerging as a prominent solution. Despite its benefits, the exact impact of quantization (from 2- to 6-bit) on the performance and accuracy of LLMs remains an active area of research. This paper investigates the performance of eight LLMs at various quantization levels, focusing on tasks such as MMLU-Pro for knowledge processing and reasoning, CRUXEval for code comprehension, and MuSR for reading comprehension. Our results show a consistent trend where higher precision (e.g., 8-bit Q8\_0) yields improved performance, albeit with diminishing returns. Aggressive quantization (e.g., 2-bit Q2\_K) usually retains acceptable accuracy, though some models show a substantial loss in performance. Our findings indicate that while lower bit precision generally reduces performance, the impact varies across models and tasks. Larger models show greater resilience to aggressive quantization, but can still undergo significant drops at lower precision levels. Mid-sized models in the 7-9 billion parameter range strike an optimal balance between efficiency and resource usage. Such results provide insights into the trade-offs between model size, quantization, and performance.