OptBench

Optimization tasks: Synthetic functions: Ackley, Levy, Schwefel

Compatible with carps

Installation

# Create conda env
conda create -n optbench python=3.12 -c conda-forge

# Activate env
conda activate optbench

# Clone repo
git clone [email protected]:automl/OptBench.git
cd OptBench

# Install 
pip install -e .

Running an optimizer from carps

The command for running an optimizer, e.g., random search from carps on OptBench looks as follows:

python -m carps.run 'hydra.searchpath=[pkg://optbench/configs]' +optimizer/randomsearch=config +problem/OptBench=synth_Ackley_2 seed=1

The breakdown of the command:

• 'hydra.searchpath=[pkg://optbench/configs]': Let hydra know where to find the configs of the OptBench package. For this, optbench needs to be installed. It is in general: pkg://PACKAGE_NAME/PATH_INSIDE_PACKAGE_TO_CONFIGS_FOLDER .
• +optimizer/randomsearch=config: select an optimizer from carps. Follows the config folder structure in carps. Beware, for other optimizers you need to install dependencies (check the repo).
• +problem/OptBench=synth_Ackley_2: Select a problem. Follows the configs folder structure in this package, starting from optbench/configs.
• seed=1: Set the seed to 1. 🙂

Of course, you can also specify the run dir etc. For more hydra overrides and parallelization, check their docs/tutorials.