Training Deep Neural Networks with 8-bit Floating Point Numbers

Open in new window