fiddle

What it is and what it does

Fiddle is a configuration library designed to let you express program parameters as Python code rather than external config files or dictionaries. It's built for machine learning workflows where deep configurability—the ability to tweak nested parameters across complex systems—is essential. The library's core idea is that configuration should be readable and maintainable, written in the same language as the rest of your codebase.

You use Fiddle by creating Config objects that represent function or class instantiations, then setting their parameters as Python attributes. This approach keeps your configuration logic explicit and debuggable. The library depends on absl-py for logging, graphviz for visualization, libcst for code manipulation, and typing-extensions for type hints. It's currently in Alpha status but actively maintained by Google, with support for Python 3.8 through 3.10.

Use it for:

• Define hyperparameter configurations for ML experiments in readable Python code instead of YAML or JSON files.
• Build nested parameter hierarchies for complex models where you need to configure sub-components at multiple levels.
• Generate and visualize configuration structures using graphviz to understand parameter dependencies.
• Manage multiple configuration variants for A/B testing or ablation studies without duplicating code.
• Share configuration logic across teams by expressing it as importable Python modules rather than static files.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Fiddle is a Python-first configuration library that lets you define and manage complex program parameters in readable Python code, particularly suited for machine learning applications.

Yes, if you're building machine learning systems or other applications that need deep, readable parameter configuration. The low install friction, active maintenance, permissive license, and zero known vulnerabilities make it a safe choice. The Alpha status is worth noting—expect the API to evolve—but the library is backed by Google and in active use. Install it if configuration-as-code in Python appeals to your workflow; skip it if you prefer external config files or simpler parameter passing.

Install

fiddle on PyPI

pip install fiddle

uv add fiddle

poetry add fiddle

Installing fiddle

Before you install

Low install friction with a pure Python wheel and four lightweight runtime dependencies. Actively maintained with recent commits; last release was in April 2024 and the repository remains active.

License in practice

Licensed under Apache 2.0 (permissive), so you can use, modify, and distribute Fiddle freely in both open-source and commercial projects without restriction.

Quickstart

pip install fiddle

import fiddle as fdl

# Define a configuration object
config = fdl.Config(some_function)
config.param_name = value

Requires Python 3.8 or later.

Verify before relying

• Whether Fiddle's configuration objects integrate with popular ML frameworks (TensorFlow, PyTorch, JAX) or if integration is manual.
• Performance characteristics when managing very large configuration hierarchies.
• Whether the library supports runtime validation or type checking of configured parameters.

Package facts

Evidence: fiddle-0.3.0-py3-none-any.whl

Keywords: fiddle, python, configuration, machine, learning

Tags

More Software Development packages