OpenSeq2Seq

OpenSeq2Seq is a TensorFlow-based toolkit for sequence-to-sequence models:

• machine translation (GNMT, Transformer, ConvS2S, …)
• speech recognition (DeepSpeech2, Wave2Letter, Jasper, …)
• speech commands (RN-50, Jasper)
• speech synthesis (Tacotron2, WaveNet…)
• language model (LSTM, …)
• sentiment analysis (SST, IMDB, …)
• image classification (ResNet-50)

Main features:

• modular architecture that allows assembling of new models from available components
• support for mixed-precision training, that utilizes Tensor Cores in NVIDIA Volta/Turing GPUs
• fast Horovod-based distributed training supporting both multi-GPU and multi-node modes

To install this toolkit, look at installation instructions. Next go to in-depth tutorials section. You can also find some useful information in the mixed precision and distributed training sections.

Reference:

1. Oleksii Kuchaev et al. Mixed-Precision Training for NLP and Speech Recognition with OpenSeq2Seq, 2018
2. Jason Li et al. Jasper: An End-to-End Convolutional Neural Acoustic Model, 2019

Disclaimer: This is a research project, not an official product by NVIDIA.