A fast parallel implementation of RNN Transducer.
This release is based on the Graves 2012 paper Sequence Transduction with Recurrent Neural Networks, both on CPU and GPU. There are bindings for Tensorflow, MXNet and PyTorch.
This release is based on the paper SPEECH RECOGNITION WITH DEEP RECURRENT NEURAL NETWORKS. The implementation is available on TensorFlow, MXNet, PyTorch.