GAMBIT tutorial

This repo collects a few simple things required for running GAMBIT, particularly in the docker image. Please let me know if you have any issues running this whatsoever!

Please clone this repo with:

git clone --recurse-submodules git@github.com:jwuerzinger/gambit_tutorial.git

to also get the main gambit repo containing example yaml files.

Setup / Installation

1. Install docker
2. Set up a python environment for analysing results:
   1. Install miniconda (anaconda works too, if you have that installed already).
   2. Make a new environment with conda create --name gambitenv --file requirements.txt
   3. Activate the environment with conda activate gambitenv

Running GAMBIT

Before running the docker image, create and empty folder to store scan results

mkdir runs

After that, run the docker image with:

Run gambit docker image with:

docker run -v $PWD/gambit/yaml_files/:/yaml_files/ -v $PWD/runs/:/runs -it --rm gambitbsm/gambit-pippi

After entering the container, you can run an example scan of the Wilson coefficients. Change the scan config to save the scan results as hdf5 with:

Printer:

  printer: hdf5
  options:
   output_file: "WC.hdf5"
   group: "/WC"

  # printer: ascii
  # options:
  #   output_file: "WC.dat"
  #   buffer_length: 100
  #   delete_file_on_restart: true

Run the example scan with:

./gambit -f /yaml_files/WC_lite.yaml

The scan will print a lot of stuff, including likelihoods and parameters at each point. You'll know that the scan was successful if it prints something like this:

Total log-likelihood: -1.3252244

Diver run finished!
ScannerBit is waiting for all MPI processes to report their shutdown condition...
Final dataset size is 7600

GAMBIT has finished successfully!

Calling MPI_Finalize...

GAMBIT will put your scan results in a local folder (runs/). This folder should look like this:

runs/WC_lite/
├── WC_lite.yaml
├── logs
│   ├── FlavBit.log
│   ├── debug.log
│   └── default.log
├── samples
│   └── WC.hdf5
└── scanner_plugins
    └── Diver
        ├── native.devo
        ├── native.raw
        └── native.rparam

Feel free to inspect all files, in particular the .log files, which contain the output logs of all parts used in the scan. Make sure you copy the contents of this folder into /runs (note the different placement of the forward slash!) before exiting the docker image: mv runs/* /runs/

Making plots

test.ipynb collects some simple ways of making plots for a scan. Use this as a draft for looking at scan results. Use the conda environment's kernel here and everything should work out-of-the box.

Make plots with pippi:

If you can make the GAMBIT plotting framework pippi work, you can make plots by simply calling:

pippi /yaml_files/WC_lite.pip

I haven't managed to do this yet though, so standalone python is fine :)

Running a scan for the MSSM:

If you want to run larger scans for the EWK pMSSM, use the yaml file:

./gambit -f /yaml_files/MSSM7.yaml

This is a very large scan though, so make sure to increase the convergence threshold and reduce the number of parallel processes:

    de:
      plugin: diver
      like: LogLike
      # NP: 19200
      # convthresh: 1e-5
      NP: 1
      convthresh: 1e-1
      verbosity: 1

Question for you: Which parts of the scan could we disable here?