On the recordRead 12 Oct 202625 facts
OpenNMT
OpenNMT · founded 2016
On the record
- API access
- Yesopennmt.net · read 22 Sep 2026
- Decoding
- Both implementations list beam search and CTranslate2 compatibility; OpenNMT-py also lists ensemble decoding, n-gram blocking, and phrase tables.opennmt.net · read 12 Oct 2026
- Framework integrations
- OpenNMT-tf documents integration with TensorFlow 2, Horovod, TensorBoard, TensorFlow Serving, and Python.github.com · read 12 Oct 2026
- Glossary tools
- Yesopennmt.net · read 22 Sep 2026
- Implementations
- OpenNMT provides OpenNMT-py, based on PyTorch, and OpenNMT-tf, based on TensorFlow.opennmt.net · read 12 Oct 2026
- Inference integration
- OpenNMT-py and OpenNMT-tf support compatibility with CTranslate2, an inference engine for Transformer models on CPU and GPU.opennmt.net · read 12 Oct 2026
- Intended users
- The FAQ describes OpenNMT as used in research and industry applications and as a toolkit for machine translation and related tasks.opennmt.net · read 12 Oct 2026
- License
- The OpenNMT ecosystem is licensed under the MIT license.opennmt.net · read 12 Oct 2026
- Model serving
- The project lists efficient model serving as a shared goal of its implementations.opennmt.net · read 12 Oct 2026
- OpenNMT-py support status
- The OpenNMT-py repository says the project is no longer actively supported and points developers to Eole, a spin-off project.github.com · read 12 Oct 2026
- Pretrained models
- The site links to pretrained models for both OpenNMT-py and OpenNMT-tf.opennmt.net · read 12 Oct 2026
- Project origin and maintainers
- OpenNMT started in December 2016 with Harvard NLP and SYSTRAN and is currently maintained by SYSTRAN and Ubiqus.opennmt.net · read 12 Oct 2026
- Purpose
- OpenNMT is an open source ecosystem for neural machine translation and neural sequence learning.opennmt.net · read 12 Oct 2026
- Security and compliance
- The official site identifies the MIT license but does not state security certifications or compliance standards.opennmt.net · read 12 Oct 2026
- Support
- The FAQ directs general questions and support requests to the community forum and says replies usually arrive within one or two work days.opennmt.net · read 12 Oct 2026
- Supported tasks
- OpenNMT supports tasks including sequence to sequence learning, language modeling, summarization, speech to text, and image to text, with availability varying by implementation and version.opennmt.net · read 12 Oct 2026
- Tokenization integration
- The ecosystem includes Tokenizer, described as a fast, customizable text tokenization library with BPE and SentencePiece support.opennmt.net · read 12 Oct 2026
- Training features
- Listed training features include data augmentation, early stopping, gradient accumulation, mixed precision, multi-GPU and multi-node training, and on-the-fly tokenization.opennmt.net · read 12 Oct 2026
