Using cogwheel with asimov¶

Asimov is able to automate cogwheel analyses using its common analysis description interface.

In order to set up a cogwheel job using asimov you’ll need to first create an asimov project, and add an event to your project which you’ll perform the analysis on.

We’ve included some quick instructions for doing this here, but you can find more a more comprehensive tutorial in the asimov documentation.

Blueprints for cogwheel¶

There are two kinds of blueprint you might want to use to configure cogwheel. The first will set project-wide defaults for the pipeline, while the second will create individual analyses. The blueprints for individual analyses can always overwrite the settings from the project-wide defaults.

First, the defaults blueprint should look something like this:

kind: configuration
pipelines:
  cogwheel:
    scheduler:
      accounting group: ligo.prod.o4.cbc.pe.bilby
      request cpus: 1
    likelihood:
      marginalization:
        distance: True
    priors:
      class: MarginalizedDistanceLVKPrior
    sampler:
      sampler kwargs:
        nlive: 1000

Save this file as cogwheel-pipeline.yaml, and you can then configure your project to default to using these values for a cogwheel analysis by running

$ asimov apply -f cogwheel-pipeline.yaml

To add a new cogwheel analysis to an event in your project you’ll need an analysis blueprint file. This should look something like this:

kind: analysis
name: cogwheel-IMRPhenomXPHM
pipeline: cogwheel
waveform:
  approximant: IMRPhenomXPHM
likelihood:
  relative binning:
    fiducial parameters:
      chirp mass: 30

Save this to a file called cogwheel.yaml. Then this can be added to the event in your project by running

$ asimov apply -f cogwheel.yaml --event GW150914

You can reuse exactly the same file for each event if you want them to use the same settings, and you only need to change the value of --event specified at the command line.