Created by: lcy-seso

Before we move to a more complicated mode for NMT task, we can implement the most basic RNN encoder-decoder without attention model and verify its convergence.