Other Apps

create_data.py

Generate training data for benchmark functions.

This script evaluates a chosen benchmark function at random or user-supplied input points and saves the resulting input/output pairs to text files. Optionally, it can also compute and save gradients.

Functions are discovered dynamically from PyTUQ’s benchmark sub-modules (bench, bench1d, bench2d, benchNd, chem, genz, poly, toy). Run python create_data.py -h to see the full list of available functions.

Outputs:

xtrain.txt — Input sample array of shape (n, d).
ytrain.txt — Output array of shape (n, m).
gtrain.txt — (only with -g ) Gradient array of shape (n, m*d).

Examples:

python create_data.py -f Ishigami -n 200 -s 0.01
python create_data.py -f Genz1 -n 500 -g -z 42
python create_data.py -f Muller-Brown -x mysamples.txt

Arguments:

Flag	Default	Description
`-n, --npts`	`100`	Number of sample points. Ignored when `-x` is given.
`-f, --func`	`Muller-Brown`	Benchmark function name (see `--help` for choices).
`-x, --xtrain`	`None`	Optional file of pre-generated input samples. When provided, the number of samples is determined by the file.
`-s, --sigma`	`0.0`	Standard deviation of additive Gaussian noise on outputs.
`-g`	off	Compute and save gradients.
`-z, --seed`	`None`	Random seed for reproducibility.

mrun.py

Utilities for parallel execution of shell tasks via multiprocessing.

This module provides helper functions for running batches of shell commands in parallel, each in its own working directory. Two strategies are offered:

mrun — spawns one multiprocessing.Process per task.
mpool — uses a multiprocessing.Pool with a fixed worker count.

When executed as a script it reads a tasks file (each line: <directory> <command>) and dispatches all tasks with mpool.

Input file: tasks — one task per line in the format <directory> <command>.

Example:

python mrun.py 4          # run tasks from ./tasks using 4 workers

Arguments:

Argument	Description
positional (`nproc`)	Number of worker processes.