How does it handle experiment metadata?

The skill implements patterns to save config snapshots, Git commit hashes, and environment information alongside your experiment results for full traceability.

What frameworks does this reproducibility skill support?

It is designed to work with standard Python research stacks, including core Python libraries, NumPy, and major machine learning frameworks like PyTorch and TensorFlow.

Can I use this for non-AI Python projects?

Yes, while it is optimized for research code, its principles of explicit configuration and state management are beneficial for any project requiring deterministic behavior.

Does this skill require heavy third-party configuration libraries?

No, it follows a lean approach, favoring explicit parameters and standard formats like YAML or TOML over complex, heavy configuration frameworks.

Python Experiment Reproducibility

Name: Python Experiment Reproducibility
Author: Silviase

bySilviase

0•

Data Science & ML

Ensures deterministic results in Python research code by centralizing configurations, controlling randomness, and recording run metadata.

This skill focuses on transforming experimental Python scripts into professional, reproducible research code. It automates the centralization of runtime parameters into structured configuration objects, implements rigorous seeding across Python and framework-specific random number generators, and establishes a systematic workflow for recording environment metadata like Git hashes and configuration snapshots. By standardizing output layouts and removing hidden global states, it ensures that every experiment run is traceable, auditable, and perfectly repeatable across different environments.

Key Features

01Deterministic seeding for Python, NumPy, and deep learning frameworks

02Standardized output directory layouts for artifacts and metrics

03Automated recording of Git hashes and environment metadata

04Elimination of implicit defaults and hidden global states

050 GitHub stars

06Centralized configuration management via YAML, TOML, or argparse

Use Cases

01Preparing machine learning research code for publication and peer review

02Debugging non-deterministic behavior in complex simulation scripts

03Standardizing experiment tracking across a distributed data science team

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add silviase/default-python-project refactoring-03-config-reproducibility

For use in Claude.ai and ChatGPT

Download Skill