Permalink
283 lines (204 sloc) 10.4 KB

Installation/Building

Dependencies

Even if you plan to install ViZDoom via PyPI or LuaRocks, you need to install some dependencies in your system first.

Linux

  • CMake 2.8+
  • Make
  • GCC 4.9+
  • Boost libraries 1.54+
  • Python 2.7+ or Python 3+ with Numpy for Python binding (optional)
  • JDK for Java binding (JAVA_HOME must be set) (optional)
  • Torch7 for Lua binding (optional)

Additionally, ZDoom dependencies are needed.

To get all dependencies (except JDK) on Ubuntu execute the following commands in the shell (requires root access):

# ZDoom dependencies
sudo apt-get install build-essential zlib1g-dev libsdl2-dev libjpeg-dev \
nasm tar libbz2-dev libgtk2.0-dev cmake git libfluidsynth-dev libgme-dev \
libopenal-dev timidity libwildmidi-dev unzip

# Boost libraries
sudo apt-get install libboost-all-dev

# Python 2 dependencies
sudo apt-get install python-dev python-pip
pip install numpy
# or install Anaconda 2 and add it to PATH

# Python 3 dependencies
sudo apt-get install python3-dev python3-pip
pip3 install numpy
# or install Anaconda 3 and add it to PATH

# Lua binding dependencies
sudo apt-get install liblua5.1-dev
# Lua shipped with Torch can be used instead, so it isn't needed if installing via LuaRocks

# Julia dependencies
sudo apt-get install julia
julia
julia> Pkg.add("CxxWrap")

To get Torch see: Getting started with Torch.

MacOS

  • CMake 2.8+
  • Clang 3.4+
  • Boost libraries
  • Python 2.7+ or Python 3+ with Numpy for Python binding (optional)
  • JDK for Java binding (JAVA_HOME must be set) (optional)
  • Torch7 for Lua binding (optional)

Additionally, ZDoom dependencies are needed.

To get dependencies install homebrew

# ZDoom dependencies and Boost libraries
brew install cmake boost sdl2 wget

# Python 2 dependencies
brew install python
pip install numpy

# Python 3 dependencies
brew install python3
pip3 install numpy
# or install

# Julia dependencies
brew cask install julia
julia
julia> Pkg.add("CxxWrap")

To get Torch see: Getting started with Torch.

Windows

  • CMake 2.8+
  • Visual Studio 2012+
  • Boost libraries
  • Python 2.7+ or Python 3.4+ with Numpy for Python binding (optional)
  • JDK for Java binding (JAVA_HOME must be set) (optional)
  • Torch7 or LUA 5.1 for Lua binding (optional)

Additionally, ZDoom dependencies are needed. Most of them are gathered in this repository: ViZDoomWinDepBin.

Installation via PyPI (recommended for Python users)

ViZDoom for Python can be installed via pip/conda on Linux and MacOS, and it is strongly recommended. However you will still need to install Linux/MacOS dependencies.

Pip installation is not supported on Windows at the moment, but we hope some day it will.

To install the most stable official release from PyPI:

# use pip3 for Python3
sudo pip install vizdoom

To install the newest version from the repository:

git clone https://github.com/mwydmuch/ViZDoom
cd ViZDoom
# use pip3 for Python3
sudo pip install .

Or without cloning it yourself:

# use pip3 for Python3
sudo pip install git+https://github.com/mwydmuch/ViZDoom

Installation via LuaRocks (recommended for Torch7 users)

ViZDoom for Torch7 can be installed via luarocks on Linux and MacOS, and it is strongly recommended. However you will still need to install Linux/MacOS dependencies.

To install the most stable official release from LuaRocks:

luarocks install vizdoom

To install the newest version from the repository:

git clone https://github.com/mwydmuch/ViZDoom
cd ViZDoom
luarocks make

Installation of Windows binaries

For Windows we are providing a compiled environment that can be download from releases page. To install it for Python, copy files to site-packages folder.

Location of site-packages depends on Python distribution:

  • Python: python_root\Lib\site-packges
  • Anaconda: anaconda_root\lib\pythonX.X\site-packages

Building

Linux

Using pip/conda is the recommended way to install ViZDoom, please try it first unless you are sure you want to compile the package by hand.

In ViZDoom's root directory:

cmake -DCMAKE_BUILD_TYPE=Release -DBUILD_PYTHON=ON -DBUILD_JAVA=ON -DBUILD_LUA=ON -DBUILD_JULIA=ON
make

where -DBUILD_PYTHON=ON, -DBUILD_JAVA=ON, -DBUILD_LUA=ON and -DBUILD_JULIA=ON CMake options for Python, Java, Lua and Julia bindings are optional (default OFF). To force building bindings for Python3 instead of the first version found use -DBUILD_PYTHON3=ON.

To build Julia binding you first need to install CxxWrap package by running julia and using Pkg.add("CxxWrap") command (see Linux dependencies). Then you need to manually set JlCxx_DIR variable:

cmake -DCMAKE_BUILD_TYPE=Release \
-DBUILD_JULIA=ON \
-DJlCxx_DIR=~/.julia/vX.X/CxxWrap/deps/usr/lib/cmake/JlCxx/

MacOS

Using pip/conda is the recommended way to install ViZDoom, please try it first unless you are sure you want to compile the package by hand.

Run CMake and build generated Makefile.

cmake -DCMAKE_BUILD_TYPE=Release -DBUILD_PYTHON=ON -DBUILD_JAVA=ON -DBUILD_LUA=ON -DBUILD_JULIA=ON
make

where -DBUILD_PYTHON=ON, -DBUILD_JAVA=ON, -DBUILD_LUA=ON and -DBUILD_JULIA=ON CMake options for Python, Java, Lua and Julia bindings are optional (default OFF). To force building bindings for Python3 instead of the first version found use -DBUILD_PYTHON3=ON.

Users with brew-installed Python/Anaconda may need to manually set PYTHON_EXECUTABLE, PYTHON_INCLUDE_DIR, PYTHON_LIBRARY variables:

It should look like this for brew-installed Python (use -DBUILD_PYTHON3=ON, include/pythonX.Xm and lib/libpythonX.Xm.dylib for Python 3):

cmake -DCMAKE_BUILD_TYPE=Release \
-DBUILD_PYTHON=ON \
-DPYTHON_EXECUTABLE=/usr/local/Cellar/python/X.X.X/Frameworks/Python.framework/Versions/X.X/bin/pythonX \
-DPYTHON_INCLUDE_DIR=/usr/local/Cellar/python/X.X.X/Frameworks/Python.framework/Versions/X.X/include/pythonX.X \
-DPYTHON_LIBRARY=/usr/local/Cellar/python/X.X.X/Frameworks/Python.framework/Versions/X.X/lib/libpythonX.X.dylib \
-DNUMPY_INCLUDES=/usr/local/Cellar/python/X.X.X/Frameworks/Python.framework/Versions/X.X/lib/pythonX.X/site-packages/numpy/core/include

Or for Anaconda (use -DBUILD_PYTHON3=ON, include/pythonX.Xm and lib/libpythonX.Xm.dylib for Python 3):

cmake -DCMAKE_BUILD_TYPE=Release \
-DBUILD_PYTHON=ON \
-DPYTHON_EXECUTABLE=~/anacondaX/bin/pythonX \
-DPYTHON_INCLUDE_DIR=~/anacondaX/include/pythonX.X \
-DPYTHON_LIBRARY=~/anacondaX/lib/libpythonX.X.dylib \
-DNUMPY_INCLUDES=~/anacondaX/lib/pythonX.X/site-packages/numpy/core/include

To build Julia binding, you first need to install CxxWrap package by running julia and using Pkg.add("CxxWrap") command (see MacOS dependencies). Then you need to manually set JlCxx_DIR variable:

cmake -DCMAKE_BUILD_TYPE=Release \
-DBUILD_JULIA=ON \
-DJlCxx_DIR=~/.julia/vX.X/CxxWrap/deps/usr/lib/cmake/JlCxx/

Windows

Setting up the compilation on Windows is really tedious so using the precompiled binaries is recommended.

vizdoom directory from Python builds contains complete Python package for Windows. You can copy it to your project directory or copy it into python_dir/lib/site-packages/vizdoom to install it globally in your system.

Run CMake GUI, select ViZDoom root directory and set paths to:

  • BOOST_ROOT
  • BOOST_INCLUDEDIR
  • BOOST_LIBRARYDIR
  • PYTHON_INCLUDE_DIR (optional, for Python/Anaconda bindings)
  • PYTHON_LIBRARY (optional, for Python/Anaconda bindings)
  • NUMPY_INCLUDES (optional, for Python/Anaconda bindings)
  • LUA_LIBRARIES (optional, for Lua/Torch bindings)
  • LUA_INCLUDE_DIR (optional, for Lua/Torch bindings)
  • ZDoom dependencies paths

In configuration select BUILD_PYTHON, BUILD_PYTHON3 and BUILD_JAVA options for Python and Java bindings (optional, default OFF).

Use generated Visual Studio solution to build all parts of ViZDoom environment.

Compilation output

Compilation output will be placed in vizdoom_root_dir/bin and it should contain following files.

  • bin/vizdoom / vizdoom.exe - ViZDoom executable
  • bin/vizdoom.pk3 - resources file used by ViZDoom (needed by ViZDoom executable)
  • bin/libvizdoom.a / vizdoom.lib - C++ ViZDoom static library
  • bin/libvizdoom.so / vizdoom.dll / libvizdoom.dylib - C++ ViZDoom dynamically linked library
  • bin/python2/vizdoom.so / vizdoom.pyd / vizdoom.dylib - ViZDoom Python 2 module
  • bin/pythonX.X/vizdoom.so / vizdoom.pyd - ViZDoom Python X.X module
  • bin/pythonX.X/pip_package - complete ViZDoom Python X.X package
  • bin/lua/vizdoom.so / vizdoom.so / vizdoom.dylib - ViZDoom Lua C module
  • bin/lua/luarocks_package - complete ViZDoom Torch package
  • bin/java/libvizdoom.so / vizdoom.dll / libvizdoom.dylib - ViZDoom library for Java
  • bin/java/vizdoom.jar - Contains ViZDoom Java classes

Manual installation

To manually install Python package copy vizdoom_root_dir/bin/pythonX.X/pip_package contents to python_root_dir/lib/pythonX.X/site-packages/site-packages/vizdoom.

To manually install Torch package copy vizdoom_root_dir/bin/lua/luarocks_package contents to torch_root_dir/install/lib/lua/5.1/vizdoom and vizdoom_root_dir/bin/lua/luarocks_shared_package contents to torch_root_dir/install/share/lua/5.1/vizdoom.