Skip to content

Latest commit

 

History

History
51 lines (44 loc) · 2.46 KB

how_to_run_QuickCW.md

File metadata and controls

51 lines (44 loc) · 2.46 KB

Quick-start Guide to QuickCW

Setup

It is advisable, to create new conda environment:

conda create --name QuickCW python=3.9

Active our new environment:

conda activate QuickCW

Install enterprise:

conda install -c conda-forge enterprise-pulsar

Clone the QuickCW repo:

git clone https://github.com/bencebecsy/QuickCW.git

Move into the repo's folder:

cd QuickCW

Install QuickCW and all remaining requirements:

pip install -e .

Running QuickCW

The main analysis code can be found in QuickCW.py, which can be executed by the wrapper script runQuickMCMC.py. Move this file into the folder where you want to run it, and modify it so that:

  • data_pkl points to the pickled pulsar object you want to analyze. Alternatively one can rewrite the script so that it loads in par/tim files. What matters in the end is that psrs contain the pulsar objects we want to use.
  • noisefile points to the json file containing the noise dictionary we plan to use for setting the white noise parameters.
  • savefile is the name of the file we want to save our results.
  • The number of iterations (N) is set properly. The example script has N=1_000_000, which is good for a quick test run to see that everything works, but not enough for an actual analysis. Depending on the details of the analysis and the dataset one might want to set at least N=100_000_000 (or even N=1_000_000_000). Note that the number of steps in the shape parameters is N/n_int_block, so for example the example script gives '1_000_000/10_000=100' steps in shape parameters.
  • If using a higher N than in the example script, it can also be useful to set thin=100 (or even 1000), which results in only saving every 100th/1000th sample to file and thus helps keep file sizes down (default is 10).

Once these are set, we can run the MCMC by executing:

python runQuickMCMC.py

If we want to run this in the background, we can execute:

nohup ./run_QuickCW.sh > nohup.out 2>&1 &

This will use run_QuickCW.sh to run the analysis in the background and send all output into the file nohup.out.

Postprocessing

Once the MCMC run finished (or even during since it saves intermediate results during runtime) all the results can be found in a single HDF5 file. Follow this jupyter notebook for a few simple postprocessing of the results, like traces, corner plots and upper limit curves: https://github.com/bencebecsy/QuickCW/blob/main/docs/plotting_results.ipynb