Skip to content
Get packs

Optuna

A Python library that automatically searches for the best settings (hyperparameters) for machine learning mode

GitHub Download v5.0.0
Stars
15k
Forks
1.5k
Licence
MIT
Language
Python
Last update
7 days ago
Type
Library

What it is

Optuna is a Python framework for hyperparameter optimization, meaning it automatically tries different settings for a program or machine learning model and keeps the combination that scores best. You write an objective function that suggests values and returns a score, and Optuna runs many trials to find good values. Its define-by-run API lets you build the search space dynamically in ordinary Python code.

  • Search for the best model type and hyperparameters for a scikit-learn regressor, as in the sample code
  • Run an optimization study with a SQLite storage file and inspect results in Optuna Dashboard
  • Scale a study to multiple workers with little or no code change, per the README's key features
  • Prune unpromising trials early to save compute

Great for

  • Machine learning practitioners tuning model settings such as learning rate or tree depth
  • Python developers who want to automate trial-and-error on any objective function
  • Teams that want to run the same search across multiple workers in parallel
  • Users who want a web dashboard to inspect optimization history and parameter importance

Not ideal for

  • People who need a no-code tool, since the README describes a Python API that requires writing code
  • Users whose objective function is very cheap and who need maximum sampling speed, where the README points to the experimental Rustuna project instead
  • Non-Python codebases, since the README only shows Python usage
Step-by-step guide: how to use OptunaInstall, first run, examples, tips and FAQs
LightGBM vs OptunaOptuna vs Weights & Biases

How to get it

  1. 1

    Install with a package manager

    You'll need: Python 3.10 or newer

    Install from PyPI

    pip install optuna

    Install dashboard (optional)

    pip install optuna-dashboard
  2. 2

    Run it

    • Install Optuna with pip install optuna
    • Write an objective function that calls trial.suggest_* methods and returns a score
    • Create a study with optuna.create_study() and call study.optimize(objective, n_trials=100)
    • Optionally, use a SQLite storage and launch optuna-dashboard sqlite:///db.sqlite3 to view results
  3. 3

    Get the most out of it

    • Use a database storage such as sqlite:///db.sqlite3 so results persist and can be viewed in the dashboard
    • Use Optuna Dashboard to view optimization history and hyperparameter importance without writing extra plotting code
    • Browse the optuna-examples repository for multi-objective, constrained, pruning, and distributed setups
    • Try the Rustuna project if sampling speed is a bottleneck, keeping in mind it is experimental

Downloads

Source code (.zip)master branch · straight from GitHubRelease v5.0.0Release notes and sourceOfficial website & docsoptuna.org

Strengths

  • Define-by-run API lets search spaces use normal Python conditionals and loops
  • Lightweight install with few requirements, per the README
  • Includes a web dashboard and visualization functions for inspecting results

Trade-offs

  • The README is truncated and does not fully document the feature set, so some details are unverified
  • Requires Python 3.10 or newer and writing Python code to use
  • Rustuna, the faster companion project, is labeled experimental and lacks some features

How it compares

ProjectStarsTypeLevelUpdatedBest for
optuna (this)15kLibraryintermediate7 days agoMachine learning practitioners tuning model settings such as learning rate or tree depth
LightGBM18.9kLibraryintermediate4 days agoData scientists working with tabular data who want a fast gradient boosting tool
nni14.4kLibraryintermediate2.3 years agoResearchers comparing hyperparameter tuning algorithms such as TPE, Hyperband, or Bayesian methods
wandb11.3kLibrarybeginner4 days agoMachine learning engineers who want to compare training runs over time
qlib49.9kLibraryadvanced7 days agoQuant researchers who want a structured ML workflow for stock data
albumentations15.3kLibrarybeginner1.4 years agoMachine learning engineers training image classification or segmentation models with limited data

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.