Build C++ code on Linux

Refer to Installation for Linux and Mac OS-X users.

System dependencies

Building BigARTM requires the following components:

  • git (any recent version) – for obtaining source code;
  • cmake (at least of version 2.8), make, g++ or clang compiler with c++11 support, boost (at least of version 1.40) – for building library and binary executable;
  • python (version 2.7 or 3.4) – for building Python API for BigARTM.

Building C++ code with CMake

BigARTM is hosted at github repository, with two branches — stable and master. stable contains latest stable release, while master is the latest version of the code. Normally master is also quite stable, so do not hesitate to try this branch.

Build is built via CMake, as in the following script.

cd ~
git clone --branch=stable
cd bigartm
mkdir build && cd build
cmake ..

Note for Linux users: By default building binary executable bigartm requiers static versions of Boost, C and C++ libraries. To alter it, run cmake command with option -DBUILD_BIGARTM_CLI_STATIC=OFF.

System-wide installation

To install command-line utility, shared library module and Python interface for BigARTM, you can type:

sudo make install

Normally this will install:

  • bigartm utility into folder /usr/local/bin/;
  • shared library (artm.dylib for Max OS-X) into folder /usr/local/lib/;
  • Python interface for BigARTM into Python-specific system directories, along with necessary dependencies.

If you want to alter target folders for binary and shared library objects, you may specify common prefix while running cmake command via option -DCMAKE_INSTALL_PREFIX=path_to_folder. By default CMAKE_INSTALL_PREFIX=/usr/local/.

Configure BigARTM Python API

Python’s interface of BigARTM is normally configured by running make install. As an alternative you may configure it manually as described below, however you still need to build native code of BigARTM with make.

# Step 1 - install Google Protobuf as dependency
# (this can be replaced by "pip install protobuf")
cd ~/bigartm/3rdparty/protobuf/python
sudo python install

# Step 2 - install Python interface for BigARTM
cd ~/bigartm/python
sudo python install

# Step 3 - point ARTM_SHARED_LIBRARY variable to (libartm.dylib) location
export ARTM_SHARED_LIBRARY=~/bigartm/build/lib/        # for linux
export ARTM_SHARED_LIBRARY=~/bigartm/build/lib/libartm.dylib     # for Mac OS X

We strongly recommend system-wide installation (e.g. make install) as there is no need to keep BigARTM code after it, so you may safely remove folder ~/bigartm/.


If you build BigARTM in existing folder build (e.g. you built BigARTM before) and encounter any errors, it may be due to out-of-date file CMakeCache.txt in folder build. In that case we strongly recommend to delete this file and try to build again.

Using BigARTM Python API you can encounter this error:

Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "build/bdist.linux-x86_64/egg/artm/wrapper/", line 19, in __init__
File "build/bdist.linux-x86_64/egg/artm/wrapper/", line 53, in _load_cdll
OSError: cannot open shared object file: No such file or directory
Failed to load artm shared library. Try to add the location of `` file into your LD_LIBRARY_PATH system variable, or to set ARTM_SHARED_LIBRARY - a specific system variable which may point to `` file, including the full path.

This error indicates that BigARTM’s python interface can not locate (libartm.dylib) files. In such case type export ARTM_SHARED_LIBRARY=path_to_artm_shared_library.

BigARTM on Travis-CI

To get a live usage example of BigARTM you may check BigARTM’s .travis.yml script and the latest continuous integration build.