Hello World!

Interface Tutorial

Requirements:

• HPC cluster

• Python 3.x

Author: DY

from pathlib import Path
import os
import time

One of the motivations behind reForge was to provide a user- and beginner-friendly interface for managing potentially hundreds or thousands of MD simulations without having to switch between multiple scripts and constantly rewriting them—all while staying within comfort of Python. This is what the ‘cli’ module is for:

from reforge.cli import run, sbatch, dojob

The idea is very simple: suppose you have “n” mutants and you need to perform “n” independent runs for each mutant to achieve sufficient sampling.

systems = [f'mutant_{i}' for i in range(4)]
mdruns = [f'mdrun_{i}' for i in range(4)]

The functions run, sbatch, and dojob allow you to execute the same script for all cases:

for system in systems:
    for mdrun in mdruns:
        run(f'echo "Hello, I\'m run {mdrun} of {system}"')

Alternatively, you can call run like this:

for system in systems:
    for mdrun in mdruns:
        run('echo', 'Hello, I\'m run', f'{mdrun} of {system}')

Alternatively, you can sbatch a script. This command passes all positional arguments to the script and keyword arguments to the SLURM scheduler.

if 0:  # Set to 1 to actually submit jobs.
    for system in systems:
        for mdrun in mdruns:
            sbatch('hello.sh', mdrun, system, mem='50M', t='00:00:01', partition='htc', qos='public')

By default, output is written to “slurm_jobs/output_file” and “slurm_jobs/error_file”.

outfiles = [f for f in Path('slurm_jobs').glob('*output*')]
for fpath in outfiles:
    with open(fpath, 'r') as file:
        # Print only the first line of each output file.
        print(file.readline(), end='')

Hello I am run mdrun_0 of system mutant_1
Hello I am run mdrun_2 of system mutant_0
Hello I am run mdrun_2 of system mutant_3
Hello I am run mdrun_1 of system mutant_2
Hello I am run mdrun_0 of system mutant_2
Hello I am run mdrun_3 of system mutant_1
Hello I am run mdrun_2 of system mutant_1
Hello I am run mdrun_1 of system mutant_3
Hello I am run mdrun_3 of system mutant_0
Hello I am run mdrun_0 of system mutant_3
Hello I am run mdrun_1 of system mutant_0
Hello I am run mdrun_1 of system mutant_1
Hello I am run mdrun_3 of system mutant_2
Hello I am run mdrun_3 of system mutant_3
Hello I am run mdrun_0 of system mutant_0
Hello I am run mdrun_2 of system mutant_2

For convenience, dojob will either submit or run the script based on its first argument. In this case, we do not want to sbatch the script.

submit = False
for system in systems:
    for mdrun in mdruns:
        dojob(submit, 'hello.sh', mdrun, system, mem='50M', t='00:00:01', partition='htc', qos='public')