OpenSeq2Seq is a toolkit for distributed and mixed precision training of sequence-to-sequence models:
- Machine translation (GNMT, Transformer, ConvS2S, …). These models have been trained with BPE vocabulary used for text tokenization.
- Speech recognition (DeepSpeech2, Wave2Letter, Jasper, …).
- Speech commands (RN-50, Jasper). Automatic speech recognition (ASR) systems can be built using a number of approaches depending on input data type, intermediate representation, model’s type and output post-processing. OpenSeq2Seq is currently focused on end-to-end CTC-based models (like original DeepSpeech model).
- Speech synthesis (Tacotron2, Tacotron2 GST, WaveNet, Centaur, …).
- Language model (LSTM with WikiText-2, LSTM with WikiText-103).
- Sentiment analysis (SST, IMDB, …).
- Image classification, a mixed precison replica of TensorFlow ResNet-50.
OpenSeq2Seq main goal is to allow researchers to most effectively explore various sequence-to-sequence models. The efficiency is achieved by fully supporting distributed and mixed-precision training.
OpenSeq2Seq is built using TensorFlow and provides all the necessary building blocks for training encoder-decoder models for neural machine translation, automatic speech recognition, speech synthesis, and language modeling.
Speech-to-text workflow uses some parts of Mozilla DeepSpeech project.
This is a research project, not an official NVIDIA product.
- Models for:
- Neural Machine Translation.
- Automatic Speech Recognition.
- Speech Synthesis.
- Language Modeling.
- NLP tasks (sentiment analysis).
- Data-parallel distributed training:
- Mixed precision training for NVIDIA Volta/Turing GPUs.
- Supports two modes for parallel training: simple multi-tower approach and Horovod-based approach.
- Supports two new optimizers: Layer-wise Adaptive Rate Control (LARC) and NovoGrad. NovoGrad is a first-order SGD-based algorithm, which computes second moments per layer instead of per weight as in Adam.
- Mixed precision with existing models.
- Interactive infer – a mode that makes it easy to demo trained models.
|New to Linux? Read our Linux for Starters series.|
|The largest compilation of the best free and open source software in the universe. Supplied with our legendary ratings charts.|
|Hundreds of in-depth reviews offering our unbiased and expert opinion on software.|
|Alternatives to Google's Products and Services examines your options to migrate from the Google ecosystem with open source Linux alternatives.|
|Alternatives to Microsoft's Products and Services recommends open source Linux software.|
|Essential Linux system tools looks at small, indispensable utilities, useful for system administrators as well as regular users.|
|Linux utilities to maximise your productivity. Small, indispensable tools, useful for anyone running a Linux machine.|
|Home computers became commonplace in the 1980s. Emulate home computers including the Commodore 64, Amiga, Atari ST, ZX81, Amstrad CPC, and ZX Spectrum.|
|Now and Then examines how promising open source software fared over the years.|
|Linux at Home looks at a range of home activities where Linux can play its part, making the most of our time at home, keeping active and engaged.|
|Linux Candy opens up to the lighter side of Linux. Have some fun!|
|Best Free Android Apps. There's a strict eligibility criteria for inclusion in this series|
|These best free books accelerate your learning of every programming language|
|These free tutorials offer the perfect tonic to the free programming books series|
|Stars and Stripes is an occasional series looking at the impact of Linux in the USA|