3ML brings together multiple instrument and fitting software packages into a common framework. Thus, installing all the pieces can be a bit of a task for the user. In order to make this a less painless process, we have packaged most of the external dependencies into conda (see below). However, if you want more control over your install, 3ML is available on PyPI via pip. If you have issues with the installs, first check that you have properly installed all the external dependencies that you plan on using. Are their libraries accessible on you system’s standard paths? If you think that you have everything setup properly and the install does not work for you, please submit an issue and we will do our best to find a solution.

XSPEC models

We allow the ability to use the models provided by `XSPEC <>`__ natively in 3ML. However, before installing `astromodels <>`__ or 3ML, you must already have XSPEC installed by one of a few methods.

HEASARC install

If you have XSPEC installed on your computer via the source code, then before installing astromodels, make sure to init your HEASARC environment. Upon installing astromodels, the process will attempt to find the proper libraries on your system if it detects XSPEC and compile the proper extensions. If you have installed your XSPEC in a non-standard directory (e.g. your home directory), the process my not find all the libraries. You can thus export an env variable that provides the explicit path to your XSPEC headers:

export XSPEC_INC_PATH=/path/to/xspec/headers

As XSPEC evolves, various models have different interfaces that depend on the version. To accommodate this, you need to set the version of XSPEC you are using with another env variable, e.g.:

export ASTRO_XSPEC_VERSION='12.12.1'


I{f using conda, it is possible to get access to XSPEC models without having to install HEASOFT. We thus recommend the following command to create your new conda environment:

conda create --name threeml -c conda-forge -c xpsecmodels python=3.7 xpsec-modelsonly numpy scipy matplotlib

This must be done before installing astromodels.

Conda installation

Conda is a platform independent package manager. It allows to install 3ML (and a lot of other software) without the need to compile anything, and in a completely separate environment from your system and your system python.

If you don’t know Conda

If you are not familiar with conda, install 3ML with the automatic script which will take care of everything:

  1. Download the script from here

  2. Run the script with bash If you plan to use XSPEC models use bash --with-xspec.

  3. The script will install 3ML and then create a script and a threeML_init.csh script. Source the former if you are using Bash (source and the second one if you are using Csh/Tcsh (source threeML_init.csh).

If you already know Conda

If you are familiar with Conda and you already have it installed, you can install 3ML by creating an environment with (note the exceptions above w.r.t. to include XSPEC support):

conda create --name threeML -c conda-forge python=3.7 numpy scipy matplotlib

then activating your environment:

conda activate threeML

NOTE: Recently a third party has uploaded a version of astromodels to the conda-forge channel. This version will break support for XSPEC currently if you used the xspec-modelsonly package. It is important to set your conda channel priority so that the threeml channel has priority over the conda-forge channel. Please verify that the astromodels installed comes from our threeml channel. We are working to resolve this hassle and we apologize to our users.

conda install -c conda-forge -c threeml astromodels threeml


If you would like to install 3ML and astromodels on their own and have more control over which dependencies you would like to use. Please to the following

  1. It is highly recommended you work within a python virtual environment to keep you base python clean

  2. install astromodels

pip install astromodels
  1. install 3ML

pip install threeml

If you need to build other dependencies such as pagmo, multinest, XSPEC, etc., it is recommended you do this before installing astromodels!


A docker with the complete 3ML environment and all dependencies can be installed with

docker pull threeml/notebook:latest

which will install a docker that can be launched with

docker run -it --rm -p 8888:8888 -v $PWD:/workdir -w /workdir threeml/notebook

which will launch a notebook server that can operate on the current directory. The notebooks and be accessed by typing localhost:8888 in your browser.

Other dependencies

You need to set up packages such as AERIE (for HAWC), or the Fermi Science Tools, before running the script, otherwise some of the functionalities will not work.

  • AERIE for HAWC: make sure that this works before running the script:

    ```bash > liff-PointSourceExpectation –version INFO [, ParseCommandLine:137]:

    liff-PointSourceExpectation Aerie version: 2.04.00 Build type: Debug

    ``` If it doesn’t, you need to set up the HAWC environment (refer to the appropriate documentation)

  • Fermi Science Tools for Fermi/LAT analysis: make sure that this works: bash > gtirfs ...  P8R2_TRANSIENT100_V6::EDISP0 P8R2_TRANSIENT100_V6::EDISP1 ... If it doesn’t, you need to configure and set up the Fermi Science Tools.

  • ROOT: ROOT is not required by 3ML, but it provides the Minuit2 minimizer which can be used in 3ML. If you have ROOT, make sure that this works before running the script:

    > root-config --version

Install from source (advanced)

Remove any previous installation you might have with:

> pip uninstall threeML
> pip uninstall astromodels
> pip uninstall cthreeML


> pip install numpy scipy ipython astropy numba cython
> pip install git+
> pip install git+ --upgrade

In order to use the HAWC plugin, you will also need to install cthreeML (run this after setting up the HAWC environment):

> pip install git+
  • NOTE: If you do not have permission to install packages in your current python environment, you can still install the packages by adding the --user option at the end of each pip command.

Tips for Mac users

The following paths need to be added to you DYLD_LIBRARY path if you have FORTRAN installed via these package managers:

  • Homebrew: DYLD_LIBRARY_PATH=/usr/local/lib/gcc/<version   number>:$DYLD_LIBRARY_PATH

  • Fink: DYLD_LIBRARY_PATH=/sw/lib/gcc<version   number>/lib:$DYLD_LIBRARY_PATH

Please inform us if you have problems related to your FORTRAN distribution.