Examples, tutorials and use cases for the Marian toolkit.
More information on https://marian-nmt.github.io
transformer-intro
-- an introduction to training your first transformer model if you are new to Marian and NMTtranslating-amun
-- examples for translating with Amuntraining-basics
-- the complete example for training a WMT16-scale modeltraining-basics-sentencepiece
-- as training-basics
, but uses built-in SentencePiece for data processing, requires Marian v1.7+transformer
-- scripts for training the transformer modelwmt2017-uedin
-- building a WMT2017-grade model for en-de based on Edinburgh's WMT2017 submissionwmt2017-transformer
-- building a better than WMT2017-grade model for en-de, beating WMT2017 submission by 1.2 BLEUwmt2017-transformer
-- building a better than WMT2017-grade model for en-de, beating WMT2017 submission by 1.2 BLEUforced-translation
-- an example showing how to apply forced translation using bilingual terminology dictionaryexample-library
-- example using Marian as a library demonstrating basic graph operations, available as a repository templateFirst download common tools:
cd tools
make all
cd ..
Next, go to the chosen directory and run run-me.sh
, e.g.:
cd training-basics
./run-me.sh
The README file in each directory provides more detailed description.
The development of Marian received funding from the European Union's Horizon 2020 Research and Innovation Programme under grant agreements 688139 (SUMMA; 2016-2019), 645487 (Modern MT; 2015-2017), 644333 (TraMOOC; 2015-2017), 644402 (HiML; 2015-2017), the European Union's Connecting Europe Facility project 2019-EU-IA-0045 (User-focused Marian; 2020-2022), the Amazon Academic Research Awards program, and the World Intellectual Property Organization.
This software contains source code provided by NVIDIA Corporation.