Cgx

Fast CUDA Grammar eXtractor for Statistical Machine Translation

View the Project on GitHub hohoCode/cgx

Cgx

Fast CUDA Grammar eXtractor for Statistical Machine Translation

Introduction

This is a GPU grammar extractor for statistical machine translation. This tool can extract hierarchical translation grammars on GPU efficiently.

For more details, please refer to the following two papers:

We will further update our codes significantly in our next release to make this tool easier to use... Stay tuned.

You are very welcome to share your usage experiences with us. Thank you.

Installation

NVCC =nvcc -arch=compute_35 -code=sm_35
CUDA_INSTALL_PATH= /opt/common/cuda/cuda-5.5.22
OPT = -O3 -I./uthash/ -I/scratch0/huah/thrust/
make
bin/strmatchcuda

Running

./bin/strmatchcuda 
./toy/hansards.f
./toy/query.f
./toy/hansards.e
./toy/hansards.a
./toy/lex.bin
gpugrammar_temp
Start Printing Gappy Phrases...

Thanks.