music-source-separation. Deep neural networks for separating singing voice from music written in TensorFlow