Dual-path RNN: efficient long sequence modeling for time-domain single-channel speech separation implemented by Pytorch - View it on GitHub