There are different ways to get scikit-learn installed:
- Install the version of scikit-learn provided by your operating system or Python distribution. This is the quickest option for those who have operating systems that distribute scikit-learn.
- Install an official release. This is the best approach for users who want a stable version number and aren’t concerned about running a slightly older version of scikit-learn.
- Install the latest development version. This is best for users who want the latest-and-greatest features and aren’t afraid of running brand-new code.
If you wish to contribute to the project, it’s recommended you install the latest development version.
Installing an official release¶
- Python (>= 2.6 or >= 3.3),
- NumPy (>= 1.6.1),
- SciPy (>= 0.9).
pip install -U scikit-learn
If there are no binary packages matching your Python version you might to try to install scikit-learn and its dependencies from Christoph Gohlke Unofficial Windows installers or from a Python distribution instead.
Scikit-learn and its dependencies are all available as wheel packages for OSX:
pip install -U numpy scipy scikit-learn
At this time scikit-learn does not provide official binary packages for Linux so you have to build from source.
Installing build dependencies¶
Installing from source requires you to have installed the scikit-learn runtime dependencies, Python development headers and a working C/C++ compiler. Under Debian-based operating systems, which include Ubuntu, you can install all these requirements by issuing:
sudo apt-get install build-essential python-dev python-setuptools \ python-numpy python-scipy \ libatlas-dev libatlas3gf-base
On recent Debian and Ubuntu (e.g. Ubuntu 13.04 or later) make sure that ATLAS is used to provide the implementation of the BLAS and LAPACK linear algebra routines:
sudo update-alternatives --set libblas.so.3 \ /usr/lib/atlas-base/atlas/libblas.so.3 sudo update-alternatives --set liblapack.so.3 \ /usr/lib/atlas-base/atlas/liblapack.so.3
In order to build the documentation and run the example code contains in this documentation you will need matplotlib:
sudo apt-get install python-matplotlib
The above installs the ATLAS implementation of BLAS (the Basic Linear Algebra Subprograms library). Ubuntu 11.10 and later, and recent (testing) versions of Debian, offer an alternative implementation called OpenBLAS.
Using OpenBLAS can give speedups in some scikit-learn modules, but can freeze joblib/multiprocessing prior to OpenBLAS version 0.2.8-4, so using it is not recommended unless you know what you’re doing.
If you do want to use OpenBLAS, then replacing ATLAS only requires a couple of commands. ATLAS has to be removed, otherwise NumPy may not work:
sudo apt-get remove libatlas3gf-base libatlas-dev sudo apt-get install libopenblas-dev sudo update-alternatives --set libblas.so.3 \ /usr/lib/openblas-base/libopenblas.so.0 sudo update-alternatives --set liblapack.so.3 \ /usr/lib/lapack/liblapack.so.3
On Red Hat and clones (e.g. CentOS), install the dependencies using:
sudo yum -y install gcc gcc-c++ numpy python-devel scipy
Building scikit-learn with pip¶
This is usually the fastest way to install or upgrade to the latest stable release:
pip install --user --install-option="--prefix=" -U scikit-learn
The --user flag ask pip to install scikit-learn in the $HOME/.local folder therefore not requiring root permission. This flag should make pip ignore any old version of scikit-learn previously installed on the system while benefitting from system packages for numpy and scipy. Those dependencies can be long and complex to build correctly from source.
The --install-option="--prefix=" flag is only required if Python has a distutils.cfg configuration with a predefined prefix= entry.
Third party distributions of scikit-learn¶
Some third-party distributions are now providing versions of scikit-learn integrated with their package-management systems.
These can make installation and upgrading much easier for users since the integration includes the ability to automatically install dependencies (numpy, scipy) that scikit-learn requires.
The following is an incomplete list of Python and OS distributions that provide their own version of scikit-learn.
Debian and derivatives (Ubuntu)¶
The Debian package is named python-sklearn (formerly python-scikits-learn) and can be installed using the following command:
sudo apt-get install python-sklearn
Additionally, backport builds of the most recent release of scikit-learn for existing releases of Debian and Ubuntu are available from the NeuroDebian repository .
A quick-‘n’-dirty way of rolling your own .deb package is to use stdeb.
Python(x,y) for Windows¶
Canopy and Anaconda for all supported platforms¶
MacPorts for Mac OSX¶
The MacPorts package is named py<XY>-scikits-learn, where XY denotes the Python version. It can be installed by typing the following command:
sudo port install py26-scikit-learn
sudo port install py27-scikit-learn
Arch Linux’s package is provided through the official repositories as python-scikit-learn for Python 3 and python2-scikit-learn for Python 2. It can be installed by typing the following command:
# pacman -S python-scikit-learn
# pacman -S python2-scikit-learn
depending on the version of Python you use.
The Fedora package is called python-scikit-learn for the Python 2 version and python3-scikit-learn for the Python 3 version. Both versions can be installed using yum:
$ sudo yum install python-scikit-learn
$ sudo yum install python3-scikit-learn
Building on windows¶
To build scikit-learn on Windows you need a working C/C++ compiler in addition to numpy, scipy and setuptools.
Picking the right compiler depends on the version of Python (2 or 3) and the architecture of the Python interpreter, 32-bit or 64-bit. You can check the Python version by running the following in cmd or powershell console:
and the architecture with:
python -c "import struct; print(struct.calcsize('P') * 8)"
The above commands assume that you have the Python installation folder in your PATH environment variable.
Once installed you should be able to build scikit-learn without any particular configuration by running the following command in the scikit-learn folder:
python setup.py install
For the 64-bit architecture, you either need the full Visual Studio or the free Windows SDKs that can be downloaded from the links below.
The Windows SDKs include the MSVC compilers both for 32 and 64-bit architectures. They come as a GRMSDKX_EN_DVD.iso file that can be mounted as a new drive with a setup.exe installer in it.
- For Python 2 you need SDK v7.0: MS Windows SDK for Windows 7 and .NET Framework 3.5 SP1
- For Python 3 you need SDK v7.1: MS Windows SDK for Windows 7 and .NET Framework 4
Both SDKs can be installed in parallel on the same host. To use the Windows SDKs, you need to setup the environment of a cmd console launched with the following flags (at least for SDK v7.0):
cmd /E:ON /V:ON /K
Then configure the build environment with:
SET DISTUTILS_USE_SDK=1 SET MSSdk=1 "C:\Program Files\Microsoft SDKs\Windows\v7.0\Setup\WindowsSdkVer.exe" -q -version:v7.0 "C:\Program Files\Microsoft SDKs\Windows\v7.0\Bin\SetEnv.cmd" /x64 /release
Finally you can build scikit-learn in the same cmd console:
python setup.py install
Replace v7.0 by the v7.1 in the above commands to do the same for Python 3 instead of Python 2.
Replace /x64 by /x86 to build for 32-bit Python instead of 64-bit Python.
The .whl package and .exe installers can be built with:
pip install wheel python setup.py bdist_wheel bdist_wininst -b doc/logos/scikit-learn-logo.bmp
The resulting packages are generated in the dist/ folder.
It is possible to use MinGW (a port of GCC to Windows OS) as an alternative to MSVC for 32-bit Python. Not that extensions built with mingw32 can be redistributed as reusable packages as they depend on GCC runtime libraries typically not installed on end-users environment.
To force the use of a particular compiler, pass the --compiler flag to the build step:
python setup.py build --compiler=my_compiler install
where my_compiler should be one of mingw32 or msvc.
See section Retrieving the latest code on how to get the development version. Then follow the previous instructions to build from source depending on your platform.
Testing scikit-learn once installed¶
Testing requires having the nose library. After installation, the package can be tested by executing from outside the source directory:
$ nosetests -v sklearn
Under Windows, it is recommended to use the following command (adjust the path to the python.exe program) as using the nosetests.exe program can badly interact with tests that use multiprocessing:
C:\Python34\python.exe -c "import nose; nose.main()" -v sklearn
This should give you a lot of output (and some warnings) but eventually should finish with a message similar to:
Ran 3246 tests in 260.618s OK (SKIP=20)
Testing scikit-learn from within the source folder¶
Scikit-learn can also be tested without having the package installed. For this you must compile the sources inplace from the source directory:
python setup.py build_ext --inplace
Test can now be run using nosetests:
nosetests -v sklearn/
This is automated by the commands:
You can also install a symlink named site-packages/scikit-learn.egg-link to the development folder of scikit-learn with:
pip install --editable .