Marian-NMT

Marian-NMT is an open-source neural machine translation framework developed by the Microsoft Translator team and the Marian community. It is written in C++ for efficiency and is optimized for training and inference of large-scale neural translation models. Marian supports both CPU and GPU execution and is widely used for research and production-grade translation systems.

Features of Marian-NMT:

• High-performance neural machine translation engine written in C++.
• Supports transformer and RNN-based architectures for flexibility in model design.
• CPU and GPU support with optimized inference and training performance.
• Subword segmentation using SentencePiece for handling rare words efficiently.
• Parallel training and multi-GPU support for large-scale datasets.
• Command-line tools for training, decoding, scoring, and vocabulary management.
• Integration with research workflows and compatibility with pre-trained Marian models.
• Open-source license, allowing customization and contribution to the library.

To Check the Version of Marian-NMT after Building from Source:

$ sudo su
$ cd /opt
$ cd marian-nmt
$ cd marian
$ cd build
$ cmake ..
$ make -j$(nproc)
$ ./marian --version

Disclaimer: Marian-NMT is an open-source software distributed under the MIT License. It is not affiliated with, endorsed by, or sponsored by any commercial entity. Marian-NMT is provided "as is," without any warranty, express or implied. Users utilize this software at their own risk. The developers and contributors are not responsible for any damages, losses, or consequences resulting from its use. Users are encouraged to review and comply with licensing terms and any applicable regulations when using Marian-NMT.