Optuna

Optuna is an open source hyperparameter optimization framework for machine learning. It automates the search for the best hyperparameter values by running many trials of a user defined objective function and learning which settings work. A search space is written directly in Python, using conditionals, loops and ordinary syntax, which the project calls eager search spaces. The framework uses modern search algorithms and prunes unpromising trials early to save compute, and searches can be parallelized across multiple threads or processes without changing the code. A study object holds the trials, and the quick start shows a complete optimization in a few lines. Optuna is framework agnostic, with examples for PyTorch, TensorFlow, Keras, scikit learn and LightGBM, so it can wrap training code written with any of them. It is aimed at machine learning engineers and researchers who want to tune models without hand written grid searches. It installs with pip and supports Python 3.10 or newer. The site links to tutorials, documentation and the GitHub repository. The project is open source and free to install. The site lists no paid plans or pricing.