Hydra Gibbs sampler

A Bayesian sampling framework for data from 21cm arrays

View project on GitHub

Example script

Hydra comes with an example script that can generate simulated data and then run an inference on it, all in one go. The script is example.py in the root directory of the Hydra repository.

For a basic example, simply run:

python example.py --ptsrc

This will run the simulation and sampler for a basic default setup with a (3, 4, 3) hexagonal grid of 10 antennas, 100 randomly-placed point sources, 30 time samples, 60 frequency channels, and 100 samples. The samples and simulated data will be output into the ./output/ directory.

To run as an MPI job, make sure that mpi4py is installed and functioning. You can then run the script like so:

mpirun -n 16 python example.py --ptsrc

At the moment, the MPI-enabled linear system solver assumes that the matrix operator can be split into square blocks, and so will only run with a square number of processes, e.g. 4, 9, 16 and so on.

Run python example.py --help to see a full list of command line options.