Reinforcement learning for a toy task of RNN decoding steps embedding. - View it on GitHub