Installing FitBenchmarking and default fitting packages
We recommend using for running/installing Fitbenchmarking. The easiest way to install FitBenchmarking is by using the Python package manager, pip.
FitBenchmarking can be installed via the command line by entering:
python -m pip install fitbenchmarking[bumps,DFO,gradient_free,minuit,SAS,numdifftools]
This will install the latest stable version of FitBenchmarking. For all available versions please visit the FitBenchmarking PyPI project. FitBenchmarking can also use additional software that cannot be installed using pip; please see Installing External Software for details.
This install will include additional optional packages – see Extra dependencies. Any of the dependencies in the square brackets can be omitted, if required, and that package will not be available for Benchmarking, or will use the version of the package already on your system, if appropriate.
Installing from source
You may instead wish to install from source, e.g., to get the very latest version of the code that is still in development.
Download this repository or clone it using git:
git clone https://github.com/fitbenchmarking/fitbenchmarking.git
Open up a terminal (command prompt) and go into the
Once you are in the right directory, we recommend that you type
python -m pip install .[bumps,DFO,gradient_free,minuit,SAS,numdifftools]
Additional software that cannot be installed via pip can also be used with FitBenchmarking. Follow the instructions at Installing External Software.
In addition to the external packages described at Installing External Software, some optional dependencies can be installed directly by FitBenchmarking. These are installed by issuing the commands
python -m pip install fitbenchmarking['option-1','option-2',...]
python -m pip install .['option-1','option-2',...]
where valid strings
bumps– installs the Bumps fitting package.
gradient_free– installs the Gradient-Free-Optimizers fitting package
levmar– installs the levmar fitting package. Note that the interface we use also requires BLAS and LAPLACK to be installed on the system, and calls to this minimizer will fail if these libraries are not present.
matlab– installs the dill module required to run matlab controllers in fitbenchmarking
minuit– installs the Minuit fitting package.
numdifftools– installs the numdifftools numerical differentiation package.