Quantizing deep convolutional networks for efficient inference: A whitepaper

摘要

We present an overview of techniques for quantizing convolutional neural networks for inference with integer weights and activations.