The rnn package provides components for implementing a wide range of Recurrent Neural Networks. It is built withing the framework of the Torch distribution for use with the nn package. The components have evolved from 3 iterations, each adding to the flexibility and capability of the package. All component modules inherit either the AbstractRecurrent or AbstractSequencer classes. Strong unit testing, continued backwards compatibility and access to supporting material are the principles followed during its development. The package is compared against existing implementations of two published papers.
Cite
@article{arxiv.1511.07889,
title = {rnn : Recurrent Library for Torch},
author = {Nicholas Léonard and Sagar Waghmare and Yang Wang and Jin-Hwa Kim},
journal= {arXiv preprint arXiv:1511.07889},
year = {2015}
}