parallel_wrapper: wrapper to sweep a parameter space
- 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)
- The single job must
- be a shell command
- write the output to a file
- Python > 3.6 (for subprocess.run and f-string)
- numpy
# put programs-to-be-wrapped into template_parallel/
# edit runner.py
python runner.py