Installation and Deployment

Standard Installation

SModelS is a Python library that requires Python version 3.6 or later. It depends on the following external Python libraries:

  • docutils>=0.3

  • numpy>=1.18.0,!=1.21.*

  • scipy>=1.0.0

  • unum>=4.0.0

  • requests>=2.0.0

  • pyslha>=3.1.0

  • pyhf>=0.6.1

  • jsonpatch>=1.25

  • jsonschema>=3.2.0

For performance reasons, we moreover recommend pytorch>=1.8.0 as backend for pyhf. This is, however, optional: if pytorch is not available, SModelS will use the default backend.

In addition, the cross section computers provided by require:

  • Pythia 8.3 (needs a C++ compiler) or Pythia 6.4.27 (needs gfortran)

  • NLL-fast 1.2 (7 TeV), 2.1 (8 TeV), and 3.1 (13 TeV) (needs a Fortran compiler)

  • Resummino (needs a C++ compiler and gfortran). Moreover, in rpm-based Linux distributions, this tool needs boost, boost-devel, gsl and gsl-devel. For deb-based distributions, libboost-dev and libgsl-dev are required.

The tools themselves, i.e. Pythia6|8, NLL-fast, and/or Resummino need not be installed separately, as the SModelS build system takes care of that (see below). SModelS however expects the tools’ dependencies (boost, gsl for Resummino, as well as the compilers) to be installed by the user. Finally, the database browser provided by requires IPython, while the interactive plotter requires plotly and pandas.

Installation Methods

  • The first installation method installs SModelS in the source directory. After downloading the source from the SModelS releases page and extracting it, run:

    make smodels

    in the top-level directory. SModelS will install the required dependencies (using pip install), but none of the external tools (Pythia6|8, NLL-fast, Resummino). If the MSSM cross section computers are needed, one can directly install SModelS together with its external tools. To this end, run:

    make smodels_externaltools

    instead of make smodels. In case the Python libraries cannot be successfully installed, the user can install them separately using his/her preferred method. Pythia, NLL-fast and Resummino can also be compiled separately running make externaltools. In case the Fortran comiler isn’t found, try make FCC=<path-to-gfortran> smodels or make FCC=<path-to-gfortran> externaltools.

  • Every external tool can also be compiled individually, run e.g.:

    make pythia6 pythia8 nllfast resummino

    Remember, though, that the compilers as well as Resummino’s dependencies (boost, gsl, see above) need to be installed already.

  • Python’s setuptools, if installed in your machine, can also be used for installing SModelS and its dependencies. After downloading the source from the SModelS releases page and extracting it, run: install

    within the main SModelS directory. If the python libraries are installed in a system folder (as is the default behavior), it will be necessary to run the install command with superuser privilege. Alternatively, one can run with the “–user” flag: install --user

    If setuptools is not installed, you can try to install the external libraries manually and then rerun For Ubuntu, SL6 machines and other platforms, a recipe is given below.

    Note that this installation method will install SModelS into the default system or user directory (e.g. ~/.local/lib/python3.10/site-packages/). Depending on your platform, the environment variables $PATH, $PYTHONPATH, $LD_LIBRARY_PATH (or $DYLD_LIBRARY_PATH) might have to be set appropriately.

    Note also, that will not attempt at downloading and compiling the external tools (Pythia6|8, NLL-fast, Resummino) at install time. Instead, this will be done on the fly, at runtime, upon call of the cross section computer(s). The external tools will also be located in the above smodels installation directory (<installdir>/lib/…).

  • Finally, SModelS is indexed on pypi. Thus, if pip3 (or pip) is installed in your machine, it is possible to install SModelS without downloading the source code:

    pip3 install smodels

    in case of system-wide installs or :

    pip3 install --user smodels

    for user-specific installations.

    This installation method will install SModelS into the default system or user directory (e.g. ~/.local/lib/python3.10/site-packages/). Depending on your platform, the environment variables $PATH, $PYTHONPATH, $LD_LIBRARY_PATH (or $DYLD_LIBRARY_PATH) might have to be set appropriately. Be aware that the example files and the parameters file discussed in the manual will also be located in your default system or user directory. Furthermore the database folder is not included (see database installation below).

    Moreover, pip will not attempt at downloading and compiling the external tools (Pythia6|8, NLL-fast, Resummino) at install time. Instead, this will be done on the fly, at runtime, upon call of the cross section computer(s). The external tools will also be located in the above smodels installation directory (<installdir>/lib/…).

    Generally, this installation method is best suited for experienced python users.

There is also a diagnostic tool available: toolbox

should list and check all internal tools (Pythia and NLL-fast) and external (numpy, scipy, unum, … ) dependencies.

In case everything fails, please contact

Installing the SModelS Database

The simplest way is to provide the URL of the official database as the database path when running SModelS (see path in parameters file). In this case the corresponding database version binary file will be automatically downloaded and used. The available database URLs can be found on the SModelS Database releases page . For using the latest official database, which is compatible with the code version used, one can also simply set:

path = official

in the parameters file. Per default, the database pickle file will be located in the users’ .cache/smodels/ directory. If you want the pickled database file to be cached in a different location, set the environment variable SMODELS_CACHEDIR accordingly, e.g. to ‘/tmp’.

For performance reasons, from v2.2.0 onwards, the signal regions (SRs) of some of the EM-type results are aggregated in the official database. (For example for CMS-SUS-19-006, the original 174 SRs have been aggregated to 40; this speeds up the calculation without too much loss in precision when combining SRs). In order to use the original, non-aggregated EM-type results, set:

path = official+nonaggregated

Alternatively, one can download the text version of the database and pickle it locally. This can be convenient if one wants to add or edit experimental results. The source code of the available databases can again be found on the SModelS Database releases page. After download, unpack it to a convenient location (e.g., to a ‘smodels-database’ folder in the SModelS source directory), and then specify the local path to this folder in the parameters file, e.g.:

path = ./smodels-database/

The first time SModelS is run, a binary file will be built using this text database folder, which can then be used in all subsequent runs. As above, by default this contains some EM-type results with aggregated SRs. The non-aggregated versions are stored as a tarball on the top level of the database folder; for v2.2.0 this is the file nonaggregated220.tar.gz. To use this, simply expand this tarball in the directory:

cd <smodels-database folder>
tar -xzvf nonaggregated220.tar.gz

The database binary file will then be re-built accordingly upon first usage.

The complete list of analyses and results included in the database can be consulted at We note that all the results in the official database release have been carefully validated and the validation material can be found at

The database can conveniently be updated independently from SModelS code updates. It suffices to unpack any new database tarball and replace the database directory or provide the path to the new folder, binary or URL address. In the same fashion, one can easily add additional results as explained below.

Adding FastLim data

The official SModelS database can be augmented with data from the fastlim results. The simplest way is to set:

path = official+fastlim

in the parameters file. It is also possible to directly download a database binary file including the fastlim maps; dedicated URLs are provied on the SModelS Database releases page for this purpose.

For using this with the text database, a tarball with the properly converted fastlim-1.0 efficiency maps (smodels-v1.1-fastlim-1.0.tgz) is located in the top level directory of the database ( it can also be downloaded separately from Github.) As for adding non-aggregated results (see above), this tarball simply needs to be exploded to be added to the database:

cd <smodels-database folder>
tar -xzvf smodels-v1.1-fastlim-1.0.tgz
rm smodels-v1.1-fastlim-1.0.tgz

Once the fastlim folders have been added to the database, SModelS auto-detects fastlim results and issues an acknowledgement.

When using the Fastlim results, please properly cite the fastlim paper; for convenience, a bibtex file is provided in the smodels-fastlim tarball.

Finally we point out that when converting the Fastlim efficiency maps efficiencies with a relative statistical uncertainty greater than 25% were set to zero. Also, per default we discard zeroes-only results.

Adding one’s own results

The Database of Experimental Results is organized as files in an ordinary directory hierarchy. Therefore, adding additional experimental results is a matter of copying and editing text files. Once the new folders and files have been added following the database structure format, SModelS automatically rebuilds the binary (Pickle) database file. The added results will then be available for using with the the SModelS tools.

System-specific Installation Instructions

Installation on Ubuntu >= 16.04

Installation on Ubuntu machines should be straightforward with superuser privileges (if you do not have superuser privileges see instructions below):

  • sudo apt install gfortran python-setuptools python-scipy python-numpy python-docutils python-argparse

  • install

Note that the last command can be run as superuser, or with the “–user” flag.

Installation on SL7

Installation on an SL7 or CentOS7 is straightforward:

  • yum install gcc-c++ scipy numpy

  • pip install unum pyslha argparse

Installation on SL6

Installation on an SL6 (Scientific Linux 6 or Scientific Linux CERN 6) machine is tricky, because SModelS requires a more recent version of scipy than is provided by SL6. We succeeded to install SModelS on SL6 by doing:

  • yum install gcc-c++ libstdc++-devel libevent-devel python-devel lapack lapack-devel blas blas-devel libgfortran python-distutils-extra

followed by:

  • pip install nose unum argparse numpy pyslha scipy

Note, that these steps can safely be done within a Python virtualenv. Pip can also be called with the “–user” flag.

Installation on other platforms or without superuser privileges using Anaconda

Another easy and platform independent way of installing SModelS without superuser priviledges is via Anaconda ( Anaconda provides a local installation of pip as well as several additional python packages. Here we assume a version of gfortran is already installed in your system.

  • download and install Anaconda for Python 3.6 (

  • make sure Anaconda’s bin and lib folders are added to your system and Python paths


and then install SModelS as a user: install --user

In order to make sure all libraries have been correctly installed, you can run: toolBox

Installation of the C++ interface

From version 1.1.1 on, SModelS comes with a simple C++ interface, see the cpp directory. Obviously, a C++ compiler is need, alongside with the python developers (header) files (libpython-dev on ubuntu, python-devel on rpm-based distros).