/mt-core

Transformer based Neural Machine Translation with Tensorflow2

Primary LanguageJavaScriptMIT LicenseMIT

mt-core

Transformer based Nerual Machine Translation

Features

  1. Based on Tensorflow2.
  2. Support gradient accumulation and mixed precision traning.
  3. Effective and efficient data pipline.
  4. Support checkpoint averaging.
  5. Beam search for inference.
  6. Support file translation: TXT, DOC, DOCX, PPTX, PDF, HTML, XML
  7. Preserve file format after translation.
  8. A common and easy-to-use Translator interface.
  9. Support several NMT model types, such as Checkpoint, SavedModel and CTranslate2.
  10. Manage NMT models through config file.
  11. RESTful service interface for translation.
  12. Provide Web interface for text and file translation.
  13. Support translation volume limit and API key.
  14. Friendly GUI for training and manipulating NMT models.

Contents

The contents of this repository are as follows:

References

  1. https://github.com/OpenNMT/OpenNMT-tf
  2. https://github.com/ymoslem/DesktopTranslator
  3. https://github.com/argosopentech/translate-html
  4. https://github.com/argosopentech/argos-translate
  5. https://github.com/LibreTranslate/LibreTranslate