NeurST News
**March 29, 2022**: Release of [GigaST dataset](/datasets/GigaST): a large-scale speech translation corpus.
**Aug 16, 2021**: Release of models and results for [IWSLT 2021 offline ST and simultaneous translation task](/examples/iwslt21).
**June 15, 2021**: Integration of [LightSeq](https://github.com/bytedance/lightseq) for training speedup, see the [experimental branch](https://github.com/bytedance/neurst/tree/lightseq).
**March 28, 2021**: The v0.1.1 release includes the instructions of weight pruning and quantization aware training for transformer models, and several more features. See the [release note](https://github.com/bytedance/neurst/releases/tag/v0.1.1) for more details.
**Dec. 25, 2020**: The v0.1.0 release includes the overall design of the code structure and recipes for training end-to-end ST models. See the [release note](https://github.com/bytedance/neurst/releases/tag/v0.1.0) for more details.
## Highlights
- **Production ready**: The model trained by NeurST can be directly exported as TF savedmodel format and use TensorFlow-serving. There is no gap between the research model and production model. Additionally, one can use [LightSeq](https://github.com/bytedance/lightseq) for NeurST model serving with a much lower latency.
- **Light weight**: NeurST is designed specifically for end-to-end ST and NMT models, with clean and simple code. It has no dependency on Kaldi, which simplifies installation and usage.
- **Extensibility and scalability**: NeurST has the careful design for extensibility and scalability. It allows users to customize `Model`, `Task`, `Dataset` etc. and combine each other.
- **High computation efficiency**: NeurST has high computation efficiency and can be further optimized by enabling mixed-precision and XLA. Fast distributed training using [`Byteps`](https://github.com/bytedance/byteps) / [`Horovod`](https://github.com/horovod/horovod) is also supported for large-scale scenarios.
- **Reliable and reproducible benchmarks**: NeurST reports strong baselines with well-designed hyper-parameters on several benchmark datasets (MT&ST). It provides a series of recipes to reproduce them.
## Pretrained Models & Performance Benchmarks
NeurST provides reference implementations of various models and benchmarks. Please see [examples](/examples) for model links and NeurST benchmark on different datasets.
- Text Translation
- [Transformer on WMT14 en->de](/examples/translation)
- Speech-to-Text Translation
- [libri-trans](/examples/speech_transformer/augmented_librispeech)
- [MuST-C](/examples/speech_transformer/must-c)
## Requirements and Installation
- Python version >= 3.6
- TensorFlow >= 2.3.0
Install NeurST from source:
```
git clone https://github.com/bytedance/neurst.git
cd neurst/
pip3 install -e .
```
If there exists ImportError during running, manually install the required packages at that time.
## Citation
```
@InProceedings{zhao2021neurst,
author = {Chengqi Zhao and Mingxuan Wang and Qianqian Dong and Rong Ye and Lei Li},
booktitle = {the 59th Annual Meeting of the Association for Computational Linguistics (ACL): System Demonstrations},
title = {{NeurST}: Neural Speech Translation Toolkit},
year = {2021},
month = aug,
}
```
## Contact
Any questions or suggestions, please feel free to contact us: [zhaochengqi.d@bytedance.com](mailto:zhaochengqi.d@bytedance.com), [wangmingxuan.89@bytedance.com](mailto:wangmingxuan.89@bytedance.com).
## Acknowledgement
We thank Bairen Yi, Zherui Liu, Yulu Jia, Yibo Zhu, Jiaze Chen, Jiangtao Feng, Zewei Sun for their kind help.