Quantizing Convolutional Neural Networks for Low-Power High-Throughput Inference Engines

TL;DR

This paper introduces a quantization scheme for convolutional neural networks that enables low-power, high-throughput inference by using more efficient arithmetic than half-precision floating-point, without retraining.

Contribution

The authors propose a calibration-based quantization method that maintains accuracy without retraining, suitable for resource-constrained inference devices.

Findings

Achieves end-to-end accuracy comparable to floating-point models

Reduces computational complexity for inference engines

Eliminates the need for retraining during quantization

Abstract

Deep learning as a means to inferencing has proliferated thanks to its versatility and ability to approach or exceed human-level accuracy. These computational models have seemingly insatiable appetites for computational resources not only while training, but also when deployed at scales ranging from data centers all the way down to embedded devices. As such, increasing consideration is being made to maximize the computational efficiency given limited hardware and energy resources and, as a result, inferencing with reduced precision has emerged as a viable alternative to the IEEE 754 Standard for Floating-Point Arithmetic. We propose a quantization scheme that allows inferencing to be carried out using arithmetic that is fundamentally more efficient when compared to even half-precision floating-point. Our quantization procedure is significant in that we determine our quantization scheme parameters by calibrating against its reference floating-point model using a single inference batch rather than (re)training and achieve end-to-end post quantization accuracies comparable to the reference model.