Skip to content

Tutorial and bird's eye view

Creating a Default Configuration and Launching the Full MEGATOP Pipeline

This will allow you to test the pipeline and see the main steps. This is however a minimal test case. For instance, more noise simulations would be needed for the ouptut to make sense.

  1. Create a configuration file using the default settings:

    python get_example_config.py

    This will save the default configuration file in ./paramfiles/default_config.yaml.

  2. Set the required paths in the config file:

    Location in YAML Description
    data_dirs.root Path where simulations are saved.
    output_dirs.root Path where pipeline outputs are saved (defaults to current directory if not changed).

    By default the config uses fiducial_cmb.compute_from_camb: true, which computes fiducial CMB spectra on the fly via CAMB. If you prefer to provide precomputed spectra files, set compute_from_camb: false and point fiducial_cmb.fiducial_lensed_scalar and fiducial_cmb.fiducial_unlensed_scalar_tensor_r1 to your FITS files (lensed scalar spectra and unlensed tensor spectra with r=1 respectively).

  3. Run the full pipeline using one of the two approaches below.

Snakemake is an optional dependency (requires Python >= 3.11), installed via pip install megatop[snake]. It handles step ordering, parallelism, and skips steps whose outputs already exist.

snakemake --cores 4 --configfile paramfiles/default_config.yaml

To preview which steps will run without executing them:

snakemake --cores 4 --configfile paramfiles/default_config.yaml --dry-run

On a SLURM allocation, replace 4 with $SLURM_CPUS_PER_TASK.

Option B — Bash script (legacy)

bash runfiles/run_default.sh

WARNING: run_default.sh expects a conda environment named megatop where Megatop is installed.

A few notes:

  • This default config generates only one sky and one noise simulation — results will not be statistically meaningful.
  • It uses nside=512, which implies somewhat slow execution and sizeable memory use. Feel free to lower general_pars.nside in the config file; general_pars.lmax must be adjusted accordingly.