Using Containers

Containers package software together, providing all the neccersery dependencies to run bilby. Using containers can help eliminate installation/dependency issues, as well as ensure things are run in a standardized, reproducible environment. For a general introduction to containers, see the docker documentation.

We provide two sorts of containers for bilby

Both of these are industry standard software, see this page for getting started with Docker and this page for getting started with singularity.

While Docker is better known throughout software development, singularity is often preferred in science applications as it focuses on reproducibility, makes interacting with your home file system easy, and is typically preferred by cluster admins as it does not require root-priveledges as run time.

Running a bilby script with Singularity

Once you have singularity installed, you can pull (or download) the bilby image by running the command

$ singularity pull shub://lscsoft/bilby:0.3.6

This will download a file lscsoft-bilby-master-0.3.6.simg, you can rename this file as you wish, for the sake of this example let’s rename it

$ mv lscsoft-bilby-master-0.3.6.simg bilby.simg

Next, we start the container interactively

$ singularity shell -B $PWD bilby.simg

This will drop you into the container with the current working directory available. So, for example, if you had a bilby script that you wished to run, simply do

$ python

from inside the container. Note, this python is not your system installed python, but that of the container. To check this, from inside the container run

$ which python
$ python --version
Python 3.7.2

The benefit of the container is that it has all the neccersery software is preinstalled, circumventing any dependency issues.


Version numbers - In the example above, we pull the 0.3.6 version of bilby. To see a list of all available versions held on singularity-hub head to the main bilby page.


Issues with interactive matplotlib backends - If you experience issues with backends, these can be caused due to the container trying to use an interactive matplotlib backend. You may wish to start the container with the command DISPLAY= singularity shell bilby.simg to force a non-interactive backend to be used.


Binding your local directory - in the examples above, we gave the flag -B $PWD. This can be neglected if you are working in your home directory.