hyperparameter_hunter

Why we included this project

If you spend more time re-running the same model configurations than actually improving them, this library is worth a look. HyperparameterHunter wraps the estimators you already use, from scikit-learn and Keras to XGBoost, LightGBM, and CatBoost, and quietly records what happens on every run. The interesting part is that it treats tuning as a persistent process rather than a one-off search: once you've logged a history of experiments, later rounds start from everything you've already tried instead of from scratch. That suits teams that want a durable record of their model work and don't want to waste effort on duplicate configurations. It also takes care of the boilerplate around cross-validation and scoring, which helps when you're juggling quick benchmark runs across several libraries.