parallel_wrapper: wrapper to sweep a parameter space

What is done

  • Make directories and copy environment
    • In each directory, some of the files can be edited.
  • Using multiprocessing.Pool and subprocess.run to run experiments in each directory
  • Reduce the results as k-dimentional np.ndarray of objects (inverse_itertools_kd_product)

Restriction

  • The single job must
    • be a shell command
    • write the output to a file

Requirements

  • Python > 3.6 (for subprocess.run and f-string)
  • numpy

Usage

# put programs-to-be-wrapped into template_parallel/
# edit runner.py
python runner.py

Note