RETURNN as a Generic Flexible Neural Toolkit with Application to Translation and Speech Recognition

05/14/2018
by   Albert Zeyer, et al.
0

We compare the fast training and decoding speed of RETURNN of attention models for translation, due to fast CUDA LSTM kernels, and a fast pure TensorFlow beam search decoder. We show that a layer-wise pretraining scheme for recurrent attention models gives over 1 allows to train deeper recurrent encoder networks. Promising preliminary results on max. expected BLEU training are presented. We are able to train state-of-the-art models for translation and end-to-end models for speech recognition and show results on WMT 2017 and Switchboard. The flexibility of RETURNN allows a fast research feedback loop to experiment with alternative architectures, and its generality allows to use it on a wide range of applications.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
09/18/2019

Espresso: A Fast End-to-end Neural Speech Recognition Toolkit

We present Espresso, an open-source, modular, extensible end-to-end neur...
research
09/27/2021

Fast-MD: Fast Multi-Decoder End-to-End Speech Translation with Non-Autoregressive Hidden Intermediates

The multi-decoder (MD) end-to-end speech translation model has demonstra...
research
09/08/2017

Training RNNs as Fast as CNNs

Common recurrent neural network architectures scale poorly due to the in...
research
05/08/2023

Fast Conformer with Linearly Scalable Attention for Efficient Speech Recognition

Conformer-based models have become the most dominant end-to-end architec...
research
07/01/2021

ESPnet-ST IWSLT 2021 Offline Speech Translation System

This paper describes the ESPnet-ST group's IWSLT 2021 submission in the ...
research
05/03/2021

On the limit of English conversational speech recognition

In our previous work we demonstrated that a single headed attention enco...
research
02/10/2021

Fast Classification Learning with Neural Networks and Conceptors for Speech Recognition and Car Driving Maneuvers

Recurrent neural networks are a powerful means in diverse applications. ...

Please sign up or login with your details

Forgot password? Click here to reset