Using a Bi-directional LSTM Model with Attention Mechanism trained on MIDI Data for Generating Unique Music.