https://github.com/jrbourbeau/dask-optuna

Scale Optuna with Dask
https://github.com/jrbourbeau/dask-optuna

Last synced: 3 months ago
JSON representation

Scale Optuna with Dask

• Host: GitHub
• URL: https://github.com/jrbourbeau/dask-optuna
• Owner: jrbourbeau
• License: mit
• Created: 2020-09-01T18:41:39.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2020-10-01T22:48:37.000Z (over 4 years ago)
• Last Synced: 2024-10-29T20:02:06.541Z (3 months ago)
• Language: Python
• Homepage: https://jrbourbeau.github.io/dask-optuna/
• Size: 5.65 MB
• Stars: 35
• Watchers: 7
• Forks: 6
• Open Issues: 6
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Dask-Optuna

[![Tests](https://github.com/jrbourbeau/dask-optuna/workflows/Tests/badge.svg)](https://github.com/jrbourbeau/dask-optuna/actions?query=workflow%3ATests+branch%3Amaster)

[![Documentation](https://github.com/jrbourbeau/dask-optuna/workflows/Documentation/badge.svg)](https://github.com/jrbourbeau/dask-optuna/actions?query=workflow%3ADocumentation+branch%3Amaster)

[![Pre-commit](https://github.com/jrbourbeau/dask-optuna/workflows/Pre-commit/badge.svg)](https://github.com/jrbourbeau/dask-optuna/actions?query=workflow%3APre-commit+branch%3Amaster)

Dask-Optuna helps improve integration between [Optuna](https://optuna.org/) and [Dask](https://dask.org/)

by leveraging Optuna's existing distributed optimization capabilities to run

optimization trials in parallel on a Dask cluster. It does this by providing a

Dask-compatible `dask_optuna.DaskStorage` storage class which wraps an

Optuna storage class (e.g. Optuna's in-memory or sqlite storage) and can be used

directly by Optuna. For example:

```python

import optuna

import joblib

import dask.distributed

import dask_optuna

def objective(trial):

    x = trial.suggest_uniform("x", -10, 10)

    return (x - 2) ** 2

with dask.distributed.Client() as client:

    # Create a study using Dask-compatible storage

    storage = dask_optuna.DaskStorage()

    study = optuna.create_study(storage=storage)

    # Optimize in parallel on your Dask cluster

    with joblib.parallel_backend("dask"):

        study.optimize(objective, n_trials=100, n_jobs=-1)

    print(f"best_params = {study.best_params}")

```

## Documentation

See the [Dask-Optuna documentation](https://jrbourbeau.github.io/dask-optuna) for more information.

## License

[MIT License](LICENSE)