Interpreting the Effects of Quantization on LLMs

TL;DR

This paper investigates how quantization affects large language models' internal representations and behavior, finding minimal impact on calibration and neuron activity, supporting quantization's reliability for model compression.

Contribution

The study provides a comprehensive interpretability analysis of quantized LLMs, revealing consistent neuron behavior and minimal impact on model calibration across different models and quantization levels.

Findings

Quantization has minor effects on model calibration.

Number of dead neurons remains stable across quantization levels.

Larger models have more salient neurons, with some exceptions.

Abstract

Quantization offers a practical solution to deploy LLMs in resource-constraint environments. However, its impact on internal representations remains understudied, raising questions about the reliability of quantized models. In this study, we employ a range of interpretability techniques to investigate how quantization affects model and neuron behavior. We analyze multiple LLMs under 4-bit and 8-bit quantization. Our findings reveal that the impact of quantization on model calibration is generally minor. Analysis of neuron activations indicates that the number of dead neurons, i.e., those with activation values close to 0 across the dataset, remains consistent regardless of quantization. In terms of neuron contribution to predictions, we observe that smaller full precision models exhibit fewer salient neurons, whereas larger models tend to have more, with the exception of Llama-2-7B. The effect of quantization on neuron redundancy varies across models. Overall, our findings suggest that effect of quantization may vary by model and tasks, however, we did not observe any drastic change which may discourage the use of quantization as a reliable model compression technique.