https://github.com/lucidfrontier45/optuna-async-helper

A Helper Library to run asynchronous optimization with Optuna
https://github.com/lucidfrontier45/optuna-async-helper

blackbox-optimization hyperparameter-tuning machinelearning optuna

Last synced: 3 months ago
JSON representation

A Helper Library to run asynchronous optimization with Optuna

• Host: GitHub
• URL: https://github.com/lucidfrontier45/optuna-async-helper
• Owner: lucidfrontier45
• License: mit
• Created: 2024-02-13T22:55:10.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2024-09-02T11:12:50.000Z (10 months ago)
• Last Synced: 2024-10-03T07:16:06.258Z (9 months ago)
• Topics: blackbox-optimization, hyperparameter-tuning, machinelearning, optuna
• Language: Python
• Homepage:
• Size: 28.3 KB
• Stars: 1
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Optuna Async Helper

A Helper Library for Optuna Async Optimization

# Install

```bash

pip install optuna-async-helper

```

# Usage

```python

import tempfile

from optuna_async_helper import (

    SearchSpace,

    SearchSpec,

    optimize,

    create_journal_storage,

    create_study,

)

def rosenbrock(x: float, y: float, z: float) -> float:

    return (z - x) ** 2 + 100 * (y - x**2) ** 2

def test_optimizer():

    search_space: SearchSpace = [

        SearchSpec(var_name="x", domain_type="float", low=-5, high=5),

        SearchSpec(var_name="y", domain_type="float", low=-5, high=5),

    ]

    z = 0.5

    initial_params = [

        {"x": 0, "y": 0},

        {"x": 1.0, "y": 0},

        {"x": 0, "y": 1.0},

    ]

    with tempfile.TemporaryDirectory() as tempdir:

        storage = create_journal_storage(f"{tempdir}/example.db")

        study = create_study(

            study_name="rosenbrock",

            storage=storage,

        )

        study = optimize(

            study,

            objective_func=rosenbrock,

            search_space=search_space,

            initial_params=initial_params,

            n_trials=10,

            batch_size=32,

            z=z,

        )

        assert study.best_value < 1.0

        assert abs(study.best_params["x"] - z) < 1.0

        assert abs(study.best_params["y"] - z) < 1.0

```

For more detail, please check `optimize` and `SearchSpec` definitions.

# Development

The project is managed by [uv](https://docs.astral.sh/uv/)