QuTiP depends on several open-source libraries for scientific computing in the Python programming language. The following packages are currently required:
| Package | Version | Details |
|---|---|---|
| Python | 2.7+ | Version 3.3+ is highly recommended. |
| Numpy | 1.7+ | Not tested on lower versions. |
| Scipy | 0.13+ | Lower versions have missing features. |
| Matplotlib | 1.2.0+ | Some plotting does not work on lower versions. |
| Cython | 0.15+ | Needed for compiling some time-dependent Hamiltonians. |
| GCC Compiler | 4.2+ | Needed for compiling Cython files. |
| Fortran Compiler | Fortran 90 | Needed for compiling the optional Fortran-based Monte Carlo solver. |
| BLAS library | 1.2+ | Optional, Linux & Mac only. Needed for installing Fortran Monte Carlo solver. |
| Mayavi | 4.1+ | Optional. Needed for using the Bloch3d class. |
| Python Headers | 2.7+ | Linux only. Needed for compiling Cython files. |
| LaTeX | TexLive 2009+ | Optional. Needed if using LaTeX in figures. |
| nose | 1.1.2+ | Optional. For running tests. |
| scikits.umfpack | 5.2.0+ | Optional. Faster (~2-5x) steady state calculations. |
As of version 2.2, QuTiP includes an optional Fortran-based Monte Carlo solver that has a substantial performance benefit when compared with the Python-based solver. In order to install this package you must have a Fortran compiler (for example gfortran) and BLAS development libraries. At present, these packages are only tested on the Linux and OS X platforms.
Often the easiest way is to install QuTiP is to use the Python package manager pip.
sudo pip install qutip
However, when installing QuTiP this way the Fortran-based Monte Carlo solver is not included. More detailed platform-dependent installation alternatives are given below.
Official releases of QuTiP are available from the download section on the project’s web pages
and the latest source code is available in our Github repository
In general we recommend users to use the latest stable release of QuTiP, but if you are interested in helping us out with development or wish to submit bug fixes, then use the latest development version from the Github repository.
Installing QuTiP from source requires that all the dependencies are satisfied. The installation of these dependencies is different on each platform, and detailed instructions for Linux (Ubuntu), Mac OS X and Windows are given below.
Regardless of platform, to install QuTiP from the source code run:
sudo python setup.py install
To also include the optional Fortran Monte Carlo solver, run:
sudo python setup.py install --with-f90mc
On Windows, omit sudo from the commands given above.
The easiest way to install QuTiP in Ubuntu (14.04 and later) is to use the QuTiP PPA
sudo add-apt-repository ppa:jrjohansson/qutip-releases
sudo apt-get update
sudo apt-get install python-qutip
A Python 3 version is also available, and can be installed using:
sudo apt-get install python3-qutip
With this method the most important dependencies are installed automatically, and when a new version of QuTiP is released it can be upgraded through the standard package management system. In addition to the required dependencies, it is also strongly recommended that you install the texlive-latex-extra package:
sudo apt-get install texlive-latex-extra
First install the required dependencies using:
sudo apt-get install python-dev cython python-setuptools python-nose
sudo apt-get install python-numpy python-scipy python-matplotlib
Then install QuTiP from source following the instructions given above.
Alternatively (or additionally), to install a Python 3 environment, use:
sudo apt-get install python3-dev cython python3-setuptools python3-nose
sudo apt-get install python3-numpy python3-scipy python3-matplotlib
and then do the installation from source using python3 instead of python.
Optional, but recommended, dependencies can be installed using:
sudo apt-get install texlive-latex-extra # recommended
sudo apt-get install mayavi2 # optional, for Bloch3d only
sudo apt-get install libblas-dev # optional, for Fortran Monte Carlo solver
sudo apt-get install gfortran # optional, for Fortran Monte Carlo solver
If you have not done so already, install the Apple Xcode developer tools from the Apple App Store. After installation, open Xcode and go to: Preferences -> Downloads, and install the ‘Command Line Tools’.
On the Mac OS, we recommended that you install the required libraries via MacPorts. After installation, the necessary “ports” for QuTiP may be installed via
sudo port install py34-scipy
sudo port install py34-matplotlib +latex
sudo port install py34-cython
sudo port install py34-ipython +notebook+parallel
Optional, but highly recommended ports include
sudo port install vtk5 +python27 #used for the Bloch3d class
sudo port install py27-mayavi #used for the Bloch3d class
Now, we want to tell OSX which Python and iPython we are going to use
sudo port select python python34
sudo port select ipython ipython34
To install QuTiP from Macports, run
sudo port install py-qutip
Finally, we want to set the macports compiler to the vanilla GCC version. From the command line type:
port select gcc
which will bring up a list of installed compilers, such as:
Available versions for gcc:
mp-gcc48
none (active)
We want to set the the compiler to the gcc4x compiler, where x is the highest number available, in this case mp-gcc48 (the “mp-” does not matter). To do this type:
sudo port select gcc mp-gcc48
Running port select again should give:
Available versions for gcc:
mp-gcc48 (active)
none
Installing QuTiP via Macports may take a long time as some or all of the QuTiP dependencies are build from source code. The advantage is that all dependencies are resolved automatically, and the result should be a consistent build.
A second option is to install the required Python packages using the SciPy Superpack. Further information on installing the superpack can be found on the SciPy Downloads page.
Finally, one can also use the Anaconda CE package to install all of the QuTiP
QuTiP is primarily developed for Unix-based platforms such as Linux an Mac OS X, but it can also be used on Windows. We have limited experience and ability to help troubleshoot problems on Windows, but the following installation steps have been reported to work:
Install the Python(X,Y) distribution (tested with version 2.7.3.1). Other Python distributions, such as Enthought Python Distribution or Anaconda CE have also been reported to work.
When installing Python(x,y), explicitly select to include the Cython package in the installation. This package is not selected by default.
Add the following content to the file C:/Python27/Lib/distutils/distutils.cfg (or create the file if it does not already exists):
[build]
compiler = mingw32
[build_ext]
compiler = mingw32
The directory where the distutils.cfg file should be placed might be different if you have installed the Python environment in a different location than in the example above.
Note
In some cases, to get the dynamic compilation of Cython code to work, it might be necessary to edit the PATH variable and make sure that C:\MinGW32-xy\bin appears either first in the PATH list, or possibly right after C:\Python27\Lib\site-packages\PyQt4. This is to make sure that the right version of the MinGW compiler is used if more than one is installed (not uncommon under Windows, since many packages are distributed and installed with their own version of all dependencies).
As of SciPy 0.14+, the umfpack linear solver routines for solving large-scale sparse linear systems have been replaced due to licensing restrictions. The default method for all sparse linear problems is now the SuperLU library. However, scipy still includes the ability to call the umfpack library via the scikits.umfpack module. In our experience, the umfpack solver is 2-5x faster than the SuperLU routines, which is a very noticeable performance increase when used for solving steady state solutions. We have an updated scikits.umfpack module available at http://github.com/nonhermitian/umfpack that can be installed to have SciPy find and use the umfpack library.
QuTiP is designed to take advantage of some of the optimized BLAS libraries that are available for NumPy. At present, this includes the OPENBLAS and MKL libraries. If NumPy is built against these libraries, then QuTiP will take advantage of the performance gained by using these optimized tools. As these libraries are multi-threaded, you can change the number of threads used in these packages by adding:
>>> import os
>>> os.environ['OPENBLAS_NUM_THREADS'] = '4'
>>> os.environ['MKL_NUM_THREADS'] = '4'
at the top of your Python script files, or iPython notebooks, and then loading the QuTiP framework. If these commands are not present, then QuTiP automatically sets the number of threads to one.
QuTiP includes a collection of built-in test scripts to verify that an installation was successful. To run the suite of tests scripts you must have the nose testing library. After installing QuTiP, leave the installation directory, run Python (or iPython), and call:
>>> import qutip.testing as qt
>>> qt.run()
If successful, these tests indicate that all of the QuTiP functions are working properly. If any errors occur, please check that your have installed all of the required modules. See the next section on how to check the installed versions of the QuTiP dependencies. If these tests still fail, then head on over to the QuTiP Discussion Board and post a message detailing your particular issue.
QuTiP includes an “about” function for viewing information about QuTiP and the important dependencies installed on your system. To view this information:
In [1]: from qutip import *
In [2]: about()
QuTiP: Quantum Toolbox in Python
Copyright (c) 2011 and later.
Paul D. Nation & Robert J. Johansson
('QuTiP Version: ', '3.0.1')
('Numpy Version: ', '1.8.0')
('Scipy Version: ', '0.13.3')
('Cython Version: ', '0.20.1post0')
('Matplotlib Version: ', '1.3.1')
('Fortran mcsolver: ', 'True')
('scikits.umfpack: ', 'False')
('Python Version: ', '2.7.6')
('Platform Info: ', 'Linux', '(x86_64)')
()