RETURNN: The RWTH Extensible Training framework for Universal Recurrent Neural Networks

TL;DR

RETURNN is an extensible, configurable training framework optimized for efficient multi-GPU training of recurrent neural networks, supporting state-of-the-art LSTM models for speech and handwriting recognition.

Contribution

It introduces a flexible, open-source software framework specifically designed for training complex RNN topologies on multiple GPUs, facilitating research and development.

Findings

Supports training of deep bidirectional LSTM models

Achieved successful results in evaluation campaigns

Enables efficient multi-GPU training of RNNs

Abstract

In this work we release our extensible and easily configurable neural network training software. It provides a rich set of functional layers with a particular focus on efficient training of recurrent neural network topologies on multiple GPUs. The source of the software package is public and freely available for academic research purposes and can be used as a framework or as a standalone tool which supports a flexible configuration. The software allows to train state-of-the-art deep bidirectional long short-term memory (LSTM) models on both one dimensional data like speech or two dimensional data like handwritten text and was used to develop successful submission systems in several evaluation campaigns.