Skip to content

Installation

ropt is distributed on PyPI and can be installed with any standard Python package manager. It requires Python 3.11 or newer.

Install the core package

Using pip:

pip install ropt

The core install includes the SciPy-based optimizer and sampler backends, which are sufficient for most basic optimization tasks.

Optional extras

ropt exposes a few optional dependency groups that enable additional functionality:

Extra Pulls in Enables
pandas pandas Exporting Results to data frames via results_to_dataframe.
hpc pysqa, cloudpickle Running evaluations on HPC clusters via HPCExecutor.
external cloudpickle Running evaluations in an external Python process via the external backend.

Install with:

pip install "ropt[pandas]"
pip install "ropt[pandas,hpc,external]"

Plugin packages

Additional optimization backends are provided as standalone packages that register themselves through Python entry points. Once installed they become available to ropt automatically:

Package Adds
ropt-dakota Algorithms from the Dakota toolkit.
ropt-nomad The MADS algorithm via NOMAD.
ropt-pymoo Algorithms from pymoo.

Install any of them alongside ropt:

pip install ropt ropt-pymoo

After installation, select a method from the plugin by setting the backend.method field in your configuration (see Configuration) to a "plugin/method" string such as "pymoo/nelder-mead".

Verifying the installation

A quick sanity check:

# Print the current version:
from ropt.version import __version__
print(__version__)

# Verify the SciPy backend is available:
from ropt.workflow import find_backend_plugin
print(find_backend_plugin("slsqp"))  # should print "scipy"

If scipy is printed, the default backend plugin is working. Any additional plugin packages you installed can be verified by checking their methods in the same way.

Where to next