A Way to Benchmark Your Deep Learning Framework On-premise

Recently, our organization(@Machine Learning Cell, SK Telecom) started actively using MXNet and Gluon. As part of the DevOps organization, we need not only to develop new models constantly but also to maintain and deploy them seamlessly, even in a disruptive environment. There are many deep learning frameworks these days: MXNet, TensorFlow, PyTorch, Theano, to name a few. But, from the "Ops" point of view, maintaining many models for all different frameworks can be challenging. TensorFlow has many high-quality reference code and tutorials.