Dragonfly

Overview

Dragonfly is an open source Python library for scalable Bayesian optimisation. It uses probabilistic models to optimize expensive black-box functions efficiently. The library supports multiple acquisition functions and parallel evaluations.

Best for

Best for
Developers needing scalable Bayesian optimisation for expensive black-box function evaluations, especially in observability and ML tuning

Use cases

• Hyperparameter tuning for machine learning models
• Optimizing simulation parameters in scientific computing
• Tuning observability system configurations

Notes

Dragonfly is an open source Python library for scalable Bayesian optimisation. It uses probabilistic models to optimize expensive black-box functions efficiently. The library supports multiple acquisition functions and parallel evaluations.

893 stars on GitHub. Last updated 2023-06-19. Licensed MIT.

Use cases

• Hyperparameter tuning for machine learning models
• Optimizing simulation parameters in scientific computing
• Tuning observability system configurations

Pros

• Scalable Bayesian optimisation with support for parallel evaluations
• Open source with active community contributions
• Integrates well with existing Python workflows

Cons

• Limited to Bayesian optimisation methods, not a general optimization library
• Requires careful selection of acquisition functions for best results
• Smaller community compared to alternatives like Optuna or Hyperopt

Indexed from awesome-llmops and enriched against its public facts.