Fast Neural Machine Translation in C++
Go to file
2016-04-18 22:47:41 +02:00
scripts Shebang 2016-04-18 14:55:25 +00:00
src Clean mblas implementation - remove uneeded files 2016-04-18 21:53:52 +02:00
.gitignore Add python's pyc files to ignore 2016-04-15 16:27:48 +02:00
CMakeLists.txt merge 2016-04-18 19:55:41 +02:00
LICENSE Initial commit 2016-04-14 11:27:41 +01:00
README.md Change README to markdown 2016-04-18 22:47:41 +02:00

amuNN

Requirements:

  • CMake
  • Boost
  • CUDA
  • KenLM

Compilation

The project is a standard Cmake project:

mkdir build cd build cmake .. make -j

KenLM

To compile amuNN you need the lastest version of kenLM and pass the path as a Cmake variable KENLM, e.g. :

cmake -DKENLM=../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=/home/marcin/myboost cmake .. -DCMAKE_CXX_COMPILER=g++-4.9 -DCUDA_HOST_COMPILER=/usr/bin/g++-4.9

Vocabularies need to be converted to text with the scripts in the scripts folder.