marian/README.md

amuNN

A C++ decoder for Neural Machine Translation (NMT) models trained with Theano-based scripts from DL4MT (https://github.com/nyu-dl/dl4mt-tutorial)

Requirements:

• CMake 3.5.1 (due to CUDA related bugs in earlier versions)
• Boost 1.5
• CUDA 7.5
• yaml-cpp 0.5 (https://github.com/jbeder/yaml-cpp.git)

Optional

• KenLM for n-gram language models (https://github.com/kpu/kenlm, current master)

Compilation

The project is a standard Cmake out-of-source build:

mkdir build
cd build
cmake ..
make -j

Or with KenLM support:

cmake .. -DKENLM=path/to/kenlm

On Ubuntu 16.04, you need g++4.9 and cuda-7.5 and a boost version compiled with g++4.9

CUDA_BIN_PATH=/usr/local/cuda-7.5 BOOST_ROOT=/path/to/custom/boost cmake .. \
-DCMAKE_CXX_COMPILER=g++-4.9 -DCUDA_HOST_COMPILER=/usr/bin/g++-4.9 -DKENLM=path/to/kenlm

Vocabularies (*.pkl extension) need to be converted to text with the scripts in the scripts folder.

python scripts/vocab2txt.py vocab.en.pkl > vocab.en