iracepy-tiny

Pythonic bindings for irace: Iterated Racing for Automatic Algorithm Configuration.

This package is a fork/reimplementation of iracepy with a reduced feature set, which tries to be somewhat more pythonic (i.e. define scenarios and parameter spaces dynamically in code, not in static configuration files).

For more information on irace, see its documentation directly.

If you need any options for irace that are not yet implemented here, feel free to open an issue or fork the repository and open a PR.

Getting Started

Requirements

• irace R package (version 4.3 from CRAN)
• Python

Tested with Python 3.12 and R 4.4.1.

Installation

pip install git+https://github.com/Saethox/iracepy-tiny#egg=irace

Usage

from irace import irace, ParameterSpace, Scenario, Experiment


def target_runner(experiment: Experiment, scenario: Scenario) -> float:
    ...


parameter_space = ParameterSpace(...)
scenario = Scenario(...)

result = irace(target_runner, parameter_space)

Examples

See the examples directory.