Improving Neural Network Quantization without Retraining using Outlier Channel Splitting

TL;DR

This paper introduces Outlier Channel Splitting (OCS), a novel method for quantizing neural networks without retraining by addressing outliers through channel duplication and splitting, improving quantization accuracy on standard hardware.

Contribution

The paper proposes OCS, a training-free outlier handling technique that enhances neural network quantization performance on commodity hardware.

Findings

OCS outperforms existing clipping methods on ImageNet classification.

OCS achieves comparable or better results with minimal overhead.

Method works effectively on language modeling tasks.

Abstract

Quantization can improve the execution latency and energy efficiency of neural networks on both commodity GPUs and specialized accelerators. The majority of existing literature focuses on training quantized DNNs, while this work examines the less-studied topic of quantizing a floating-point model without (re)training. DNN weights and activations follow a bell-shaped distribution post-training, while practical hardware uses a linear quantization grid. This leads to challenges in dealing with outliers in the distribution. Prior work has addressed this by clipping the outliers or using specialized hardware. In this work, we propose outlier channel splitting (OCS), which duplicates channels containing outliers, then halves the channel values. The network remains functionally identical, but affected outliers are moved toward the center of the distribution. OCS requires no additional training and works on commodity hardware. Experimental evaluation on ImageNet classification and language modeling shows that OCS can outperform state-of-the-art clipping techniques with only minor overhead.