Quantizing deep convolutional networks for efficient inference: A whitepaper