Installation

$ conda install -c conda-forge bilby

Supported python versions: 3.9-3.11.

This will install all requirements for running bilby for general inference problems, including our default sampler dynesty. Other samplers will need to be installed via pip or the appropriate means.

To install the requirements for running bilby.gw for gravitational wave inference, please additionally run the following commands.

$ conda install -c conda-forge gwpy python-lalsimulation

Install bilby from source

bilby is developed and tested with Python 3.9-3.11. In the following, we assume you have a working python installation, python pip, and git. See Installing python for our advise on installing anaconda python.

Clone the repository, install the requirements, and then install the software:

$ git clone git@git.ligo.org:lscsoft/bilby.git
$ cd bilby/
$ pip install -r requirements.txt
$ pip install .

Once you have run these steps, you have bilby installed. You can now try to run the examples.

Note

The bilby version is determined by the tags in the git repository. Therefore, you may need to run

git fetch --all --tags

to fetch the tags so that when you install from source your version information is up to date.

Note

If you do not have a git.ligo account, and receive an error message:

git@git.ligo.org: Permission denied (publickey,gssapi-keyex,gssapi-with-mic).
fatal: Could not read from remote repository.

Then you need to use the HTTPS URL, e.g., replace the first line above with

$ git clone https://git.ligo.org/lscsoft/bilby.git

Note

You may be use to using $ python setup.py install to install software from source. While it is possible to do this, current recommentations from the python community (see, e.g. this site are to use pip. For development install, the -e flag can be used.

Installing optional requirements

The requirements.txt is a minimal set of requirements for using bilby. Additionally, we provide:

1. The file optional_requirements.txt which you should install if you plan to use bilby for gravitational-wave data analysis.

2. By default, we only require the default sampler dynesty be installed. For instructions on installing other samplers, see Installing samplers.

Installing python

Most computers/clusters have a system installed python version. You may choose to use this, but here we describe an alternative. In particular, how to install the anaconda distribution python package. Firstly, download the install file, you can do this from the link above, or run the command

$ wget https://repo.anaconda.com/archive/Anaconda3-5.2.0-Linux-x86_64.sh

this will download a linux installer for python, for other versions check the anaconda page. Then, run the command

$ bash Anaconda3-5.2.0-Linux-x86_64.sh

and follow the prompts on the install screen. After this process, you should have a directory ~/anaconda3 in your home directory. This contains your python installation. In particular, if you run the command

$ which python
/home/users/USER/anaconda3/bin/python

The output here (with a suitable replacement of the path) indicates that you are using the anaconda install of python. If instead, the output says something like :code;`/usr/bin/python`, then this is not the anaconda installation, but instead the system python.

If you are finding that you have run the above steps, but python is not pointing to your anaconda install, make sure that (a) you have appended a line like this to your .bashrc file

export PATH="${HOME}/anaconda3/bin:$PATH"

and (b) that you have restarted bash for this line to take effect (i.e., run $ bash).

Note

Using your own installation of python has several advantages: its generally easier to debug, avoids conflicts with other packages, and if you end up with a broken installation you can just delete the directory and start again.

Installing lalsuite

For many gravitational-wave related examples you’ll need access to the lalsuite tool set. These should be installed by default in the requirements. However, if you have problems with your installation, you can first try to install the simple way:

$ pip install lalsuite.

If this doesn’t work, or you prefer to, you can also install from source.

First, head to https://git.ligo.org/lscsoft/lalsuite to check you have an account and SSH keys set up. Then,

$ git lfs install # you may need to install git-lfs first
$ git clone git@git.ligo.org:lscsoft/lalsuite.git
$ cd lalsuite
$ ./00boot
$ ./configure --prefix=${HOME}/lalsuite-install --disable-all-lal --enable-swig  --enable-lalsimulation
$ make; make install

Warning: in the configure line here, we have disabled everything except lalsimulation. If you need other modules, see ./configure –help.

Common problems: multiple bilby versions

Occasionally, it is possible to end up with multiple bilby installations. To diagnose this, run the following

$ python -c 'import bilby; print(bilby.__file__)'
/home/user1/anaconda3/lib/python3.6/site-packages/bilby-0.2.10py3.6.egg/bilby/__init__.pyc

Here we can see it’s installed under site-packages within anaconda. Your installation path may differ. If you believe you have multiple installs, you can verify this by running ls in that directory. Then, you can purge your installation of bilby by running

$ rm -r /home/user1/anaconda/lib/python3.6/site-packages/bilby*

Replacing the path with the output of the previous command.