{"id":18887122,"url":"https://github.com/thunlp-mt/thumt","last_synced_at":"2025-04-12T23:32:45.849Z","repository":{"id":22038942,"uuid":"94881500","full_name":"THUNLP-MT/THUMT","owner":"THUNLP-MT","description":"An open-source neural machine translation toolkit developed by Tsinghua Natural Language Processing Group","archived":false,"fork":false,"pushed_at":"2022-04-26T07:01:46.000Z","size":4445,"stargazers_count":709,"open_issues_count":24,"forks_count":198,"subscribers_count":34,"default_branch":"master","last_synced_at":"2025-04-04T02:09:56.684Z","etag":null,"topics":["deep-learning","machine-translation","neural-machine-translation"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"bsd-3-clause","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/THUNLP-MT.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null}},"created_at":"2017-06-20T10:56:45.000Z","updated_at":"2025-04-03T17:28:37.000Z","dependencies_parsed_at":"2022-08-07T10:01:28.490Z","dependency_job_id":null,"html_url":"https://github.com/THUNLP-MT/THUMT","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/THUNLP-MT%2FTHUMT","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/THUNLP-MT%2FTHUMT/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/THUNLP-MT%2FTHUMT/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/THUNLP-MT%2FTHUMT/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/THUNLP-MT","download_url":"https://codeload.github.com/THUNLP-MT/THUMT/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":248647257,"owners_count":21139081,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["deep-learning","machine-translation","neural-machine-translation"],"created_at":"2024-11-08T07:34:18.825Z","updated_at":"2025-04-12T23:32:45.831Z","avatar_url":"https://github.com/THUNLP-MT.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# THUMT: An Open Source Toolkit for Neural Machine Translation\n\n## Contents\n\n* [Introduction](#introduction)\n* [Online Demo](#online-demo)\n* [Implementations](#implementations)\n* [Notable Features](#notable-features)\n* [Documentation](#documentation)\n* [License](#license)\n* [Citation](#citation)\n* [Development Team](#development-team)\n* [Contact](#contact)\n* [Derivative Repositories](#derivative-repositories)\n\n## Introduction\n\nMachine translation is a natural language processing task that aims to translate natural languages using computers automatically. Recent several years have witnessed the rapid development of end-to-end neural machine translation, which has become the new mainstream method in practical MT systems.\n\nTHUMT is an open-source toolkit for neural machine translation developed by [the Natural Language Processing Group at Tsinghua University](http://nlp.csai.tsinghua.edu.cn/site2/index.php?lang=en). The website of THUMT is: [http://thumt.thunlp.org/](http://thumt.thunlp.org/).\n\n## Online Demo\n\nThe online demo of THUMT is available at [http://translate.thumt.cn/](http://101.6.5.207:3892/). The languages involved include Ancient Chinese, Arabic, Chinese, English, French, German, Indonesian, Japanese, Portuguese, Russian, and Spanish.\n\n## Implementations\n\nTHUMT has currently three main implementations:\n\n* [THUMT-PyTorch](https://github.com/thumt/THUMT): a new implementation developed with [PyTorch](https://github.com/pytorch/pytorch). It implements the Transformer model (**Transformer**) ([Vaswani et al., 2017](https://arxiv.org/abs/1706.03762)).\n\n* [THUMT-TensorFlow](https://github.com/thumt/THUMT/tree/tensorflow): an implementation developed with [TensorFlow](https://github.com/tensorflow/tensorflow). It implements the sequence-to-sequence model (**Seq2Seq**) ([Sutskever et al., 2014](https://papers.nips.cc/paper/5346-sequence-to-sequence-learning-with-neural-networks.pdf)), the standard attention-based model (**RNNsearch**) ([Bahdanau et al., 2014](https://arxiv.org/pdf/1409.0473.pdf)), and the Transformer model (**Transformer**) ([Vaswani et al., 2017](https://arxiv.org/abs/1706.03762)).\n\n* [THUMT-Theano](https://github.com/thumt/THUMT/tree/theano): the original project developed with [Theano](https://github.com/Theano/Theano), which is no longer updated because MLA put an end to [Theano](https://github.com/Theano/Theano). It implements the standard attention-based model (**RNNsearch**) ([Bahdanau et al., 2014](https://arxiv.org/pdf/1409.0473.pdf)), minimum risk training (**MRT**) ([Shen et al., 2016](http://nlp.csai.tsinghua.edu.cn/~ly/papers/acl2016_mrt.pdf)) for optimizing model parameters with respect to evaluation metrics, semi-supervised training (**SST**) ([Cheng et al., 2016](http://nlp.csai.tsinghua.edu.cn/~ly/papers/acl2016_semi.pdf)) for exploiting monolingual corpora to learn bi-directional translation models, and layer-wise relevance propagation (**LRP**) ([Ding et al., 2017](http://nlp.csai.tsinghua.edu.cn/~ly/papers/acl2017_dyz.pdf)) for visualizing and anlayzing RNNsearch.\n\nThe following table summarizes the features of three implementations:\n\n| Implementation | Model | Criterion | Optimizer | LRP |\n| :------------: | :---: | :--------------: | :--------------: | :----------------: |\n| Theano       |  RNNsearch | MLE, MRT, SST | SGD, AdaDelta, Adam | RNNsearch |\n| TensorFlow   |  Seq2Seq, RNNsearch, Transformer | MLE| Adam | RNNsearch, Transformer |\n| PyTorch | Transformer | MLE | SGD, Adadelta, Adam | N.A. |\n\nWe recommend using [THUMT-PyTorch](https://github.com/thumt/THUMT) or [THUMT-TensorFlow](https://github.com/thumt/THUMT/tree/tensorflow), which delivers better translation performance than [THUMT-Theano](https://github.com/thumt/THUMT/tree/theano). We will keep adding new features to [THUMT-PyTorch](https://github.com/thumt/THUMT) and [THUMT-TensorFlow](https://github.com/thumt/THUMT/tree/tensorflow).\n\n## Notable Features\n\n* Transformer ([Vaswani et al., 2017](https://arxiv.org/abs/1706.03762))\n* Multi-GPU training \u0026 decoding\n* Multi-worker distributed training\n* Mixed precision training \u0026 decoding\n* Model ensemble \u0026 averaging\n* Gradient aggregation\n* TensorBoard for visualization\n\n## Documentation\n\nThe documentation of PyTorch implementation is avaiable at [here](docs/index.md).\n\n## License\n\nThe source code is dual licensed. Open source licensing is under the [BSD-3-Clause](https://opensource.org/licenses/BSD-3-Clause), which allows free use for research purposes. For commercial licensing, please email [thumt17@gmail.com](mailto:thumt17@gmail.com).\n\n## Citation\n\nPlease cite the following paper:\n\n\u003e Zhixing Tan, Jiacheng Zhang, Xuancheng Huang, Gang Chen, Shuo Wang, Maosong Sun, Huanbo Luan, Yang Liu. [THUMT: An Open Source Toolkit for Neural Machine Translation](https://www.aclweb.org/anthology/2020.amta-research.11/). AMTA 2020.\n\n\u003e Jiacheng Zhang, Yanzhuo Ding, Shiqi Shen, Yong Cheng, Maosong Sun, Huanbo Luan, Yang Liu. 2017. [THUMT: An Open Source Toolkit for Neural Machine Translation](https://arxiv.org/abs/1706.06415). arXiv:1706.06415.\n\n## Development Team\n\nProject leaders: [Maosong Sun](http://www.thunlp.org/site2/index.php/zh/people?id=16), [Yang Liu](http://nlp.csai.tsinghua.edu.cn/~ly/), Huanbo Luan\n\nProject members:\n\nTheano: Jiacheng Zhang, Yanzhuo Ding, Shiqi Shen, Yong Cheng\n\nTensorFlow: Zhixing Tan, Jiacheng Zhang, Xuancheng Huang, Gang Chen, Shuo Wang, Zonghan Yang\n\nPyTorch: Zhixing Tan, Gang Chen\n\n## Contact\n\nIf you have questions, suggestions and bug reports, please email [thumt17@gmail.com](mailto:thumt17@gmail.com).\n\n## Derivative Repositories\n\n* [UCE4BT](https://github.com/THUNLP-MT/UCE4BT) (Improving Back-Translation with Uncertainty-based Confidence Estimation)\n* [L2Copy4APE](https://github.com/THUNLP-MT/L2Copy4APE) (Learning to Copy for Automatic Post-Editing)\n* [Document-Transformer](https://github.com/THUNLP-MT/Document-Transformer) (Improving the Transformer Translation Model with Document-Level Context)\n* [PR4NMT](https://github.com/THUNLP-MT/PR4NMT) (Prior Knowledge Integration for Neural Machine Translation using Posterior Regularization)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fthunlp-mt%2Fthumt","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fthunlp-mt%2Fthumt","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fthunlp-mt%2Fthumt/lists"}