index.rst 5.48 KB
Newer Older
Minjie Wang's avatar
Minjie Wang committed
1
Install DGL
2
===========
Minjie Wang's avatar
Minjie Wang committed
3

4
This topic explains how to install DGL. We recommend installing DGL by using ``conda`` or ``pip``.
Minjie Wang's avatar
Minjie Wang committed
5

Gan Quan's avatar
Gan Quan committed
6
7
System requirements
-------------------
8
DGL works with the following operating systems:
Gan Quan's avatar
Gan Quan committed
9
10
11

* Ubuntu 16.04
* macOS X
12
* Windows 10
Gan Quan's avatar
Gan Quan committed
13

14
15
DGL requires Python version 3.5 or later. Python 3.4 or earlier is not
tested. Python 2 support is coming.
Gan Quan's avatar
Gan Quan committed
16

17
18
DGL supports multiple tensor libraries as backends, e.g., PyTorch, MXNet. For requirements on backends and how to select one, see
`Working with different backends`_.
Gan Quan's avatar
Gan Quan committed
19

20
21
22
Starting at version 0.3, DGL is separated into CPU and CUDA builds.  The builds share the
same Python package name. If you install DGL with a CUDA 9 build after you install the
CPU build, then the CPU build is overwritten.
23

Gan Quan's avatar
Gan Quan committed
24
25
Install from conda
----------------------
26
27
If ``conda`` is not yet installed, get either `miniconda <https://conda.io/miniconda.html>`_ or
the full `anaconda <https://www.anaconda.com/download/>`_.
Gan Quan's avatar
Gan Quan committed
28

29
30
31
32
With ``conda`` installed, you will want install DGL into Python 3.5 ``conda`` environment.
Run `conda create -n dgl python=3.5` to create the environment.
Activate the environment by running `source activate dgl`.
After the ``conda`` environment is activated, run one of the following commands.
Minjie Wang's avatar
Minjie Wang committed
33
34
35

.. code:: bash

36
37
38
   conda install -c dglteam dgl              # For CPU Build
   conda install -c dglteam dgl-cuda9.0      # For CUDA 9.0 Build
   conda install -c dglteam dgl-cuda10.0     # For CUDA 10.0 Build
39
40
   conda install -c dglteam dgl-cuda10.1     # For CUDA 10.1 Build

Minjie Wang's avatar
Minjie Wang committed
41

Gan Quan's avatar
Gan Quan committed
42
43
Install from pip
----------------
44
For CPU builds, run the following command to install with ``pip``.
Minjie Wang's avatar
Minjie Wang committed
45
46
47

.. code:: bash

Gan Quan's avatar
Gan Quan committed
48
   pip install dgl
49
   
50
For CUDA builds, run one of the following commands and specify the CUDA version.
51
52
53

.. code:: bash

54
55
56
   pip install dgl           # For CPU Build
   pip install dgl-cu90      # For CUDA 9.0 Build
   pip install dgl-cu92      # For CUDA 9.2 Build
57
58
   pip install dgl-cu100     # For CUDA 10.0 Build   
   pip install dgl-cu101     # For CUDA 10.1 Build
59

60
For the most current nightly build from master branch, run one of the following commands.
61
62
63

.. code:: bash

64
65
66
67
   pip install --pre dgl           # For CPU Build
   pip install --pre dgl-cu90      # For CUDA 9.0 Build
   pip install --pre dgl-cu92      # For CUDA 9.2 Build
   pip install --pre dgl-cu100     # For CUDA 10.0 Build
68
   pip install --pre dgl-cu101     # For CUDA 10.1 Build
69

Gan Quan's avatar
Gan Quan committed
70

71
72
.. _install-from-source:

Gan Quan's avatar
Gan Quan committed
73
74
Install from source
-------------------
75
Download the source files from GitHub.
Gan Quan's avatar
Gan Quan committed
76
77

.. code:: bash
Minjie Wang's avatar
Minjie Wang committed
78

79
   git clone --recurse-submodules https://github.com/dmlc/dgl.git
Minjie Wang's avatar
Minjie Wang committed
80

81
(Optional) Clone the repository first, and then run the following:
Minjie Wang's avatar
Minjie Wang committed
82
83
84

.. code:: bash

85
   git submodule update --init --recursive
Gan Quan's avatar
Gan Quan committed
86
87
88

Linux
`````
Minjie Wang's avatar
Minjie Wang committed
89

90
Install the system packages for building the shared library. For Debian and Ubuntu
Gan Quan's avatar
Gan Quan committed
91
users, run:
Minjie Wang's avatar
Minjie Wang committed
92
93
94

.. code:: bash

Gan Quan's avatar
Gan Quan committed
95
   sudo apt-get update
Da Zheng's avatar
Da Zheng committed
96
   sudo apt-get install -y build-essential python3-dev make cmake
Minjie Wang's avatar
Minjie Wang committed
97

Gan Quan's avatar
Gan Quan committed
98
For Fedora/RHEL/CentOS users, run:
Minjie Wang's avatar
Minjie Wang committed
99

Gan Quan's avatar
Gan Quan committed
100
101
102
.. code:: bash

   sudo yum install -y gcc-c++ python3-devel make cmake
Minjie Wang's avatar
Minjie Wang committed
103

104
105
106
Build the shared library. Use the configuration template ``cmake/config.cmake``.
Copy it to either the project directory or the build directory and change the
configuration as you wish. For example, change ``USE_CUDA`` to ``ON`` will
107
enable a CUDA build. You could also pass ``-DKEY=VALUE`` to the cmake command
108
109
for the same purpose.

110
- CPU-only build
111
112
113
114
115
116
   .. code:: bash

      mkdir build
      cd build
      cmake ..
      make -j4
117
- CUDA build
118
119
120
121
122
123
124
125
   .. code:: bash

      mkdir build
      cd build
      cmake -DUSE_CUDA=ON ..
      make -j4

Finally, install the Python binding.
Minjie Wang's avatar
Minjie Wang committed
126
127
128

.. code:: bash

Gan Quan's avatar
Gan Quan committed
129
130
131
132
133
134
   cd ../python
   python setup.py install

macOS
`````

135
Installation on macOS is similar to Linux. But macOS users need to install build tools like clang, GNU Make, and cmake first. These installation steps were tested on macOS X with clang 10.0.0, GNU Make 3.81, and cmake 3.13.1.
Minjie Wang's avatar
Minjie Wang committed
136

Gan Quan's avatar
Gan Quan committed
137
Tools like clang and GNU Make are packaged in **Command Line Tools** for macOS. To
138
install, run the following:
Minjie Wang's avatar
Minjie Wang committed
139
140
141

.. code:: bash

Gan Quan's avatar
Gan Quan committed
142
   xcode-select --install
Minjie Wang's avatar
Minjie Wang committed
143

Gan Quan's avatar
Gan Quan committed
144
To install other needed packages like cmake, we recommend first installing
145
**Homebrew**, which is a popular package manager for macOS. To learn more, see the `Homebrew website <https://brew.sh/>`_.
Gan Quan's avatar
Gan Quan committed
146

147
After you install Homebrew, install cmake.
Minjie Wang's avatar
Minjie Wang committed
148
149
150

.. code:: bash

Gan Quan's avatar
Gan Quan committed
151
   brew install cmake
Minjie Wang's avatar
Minjie Wang committed
152

153
154
Go to root directory of the DGL repository, build a shared library, and
install the Python binding for DGL.
Gan Quan's avatar
Gan Quan committed
155
156

.. code:: bash
Minjie Wang's avatar
Minjie Wang committed
157

Gan Quan's avatar
Gan Quan committed
158
159
   mkdir build
   cd build
160
   cmake -DUSE_OPENMP=off ..
Gan Quan's avatar
Gan Quan committed
161
162
163
   make -j4
   cd ../python
   python setup.py install
Minjie Wang's avatar
Minjie Wang committed
164

Gan Quan's avatar
Gan Quan committed
165
166
Windows
```````
Minjie Wang's avatar
Minjie Wang committed
167

168
The Windows source build is tested with CMake and MinGW/GCC.  We highly recommend
Gan Quan's avatar
Gan Quan committed
169
using CMake and GCC from `conda installations <https://conda.io/miniconda.html>`_.  To
170
get started, run the following:
Minjie Wang's avatar
Minjie Wang committed
171
172
173

.. code:: bash

Gan Quan's avatar
Gan Quan committed
174
175
   conda install cmake m2w64-gcc m2w64-make

176
Build the shared library and install the Python binding.
Gan Quan's avatar
Gan Quan committed
177
178
179
180
181

.. code::

   md build
   cd build
182
   cmake -DCMAKE_CXX_FLAGS="-DDMLC_LOG_STACK_TRACE=0 -DDGL_EXPORTS" -DCMAKE_MAKE_PROGRAM=mingw32-make .. -G "MSYS Makefiles"
Gan Quan's avatar
Gan Quan committed
183
184
185
   mingw32-make
   cd ..\python
   python setup.py install
Quan (Andy) Gan's avatar
Quan (Andy) Gan committed
186

187
You can also build DGL with MSBuild.  With `MS Build Tools <https://go.microsoft.com/fwlink/?linkid=840931>`_
Quan (Andy) Gan's avatar
Quan (Andy) Gan committed
188
and `CMake on Windows <https://cmake.org/download/>`_ installed, run the following
189
in VS2017 x64 Native tools command prompt.
Quan (Andy) Gan's avatar
Quan (Andy) Gan committed
190
191
192
193
194
195
196
197
198

.. code::

   MD build
   CD build
   cmake -DCMAKE_CXX_FLAGS="/DDGL_EXPORTS" -DCMAKE_CONFIGURATION_TYPES="Release" .. -G "Visual Studio 15 2017 Win64"
   msbuild dgl.sln
   cd ..\python
   python setup.py install