mlvern is a lightweight Python framework for building reproducible and well-organized machine learning workflows. It provides clear tooling for dataset management, experiment tracking, model versioning, and evaluation reporting.
Project documentation: https://ml-vern.readthedocs.io/en/latest/
Machine learning projects often become difficult to maintain due to scattered datasets, untracked experiments, and inconsistent model artifacts. mlvern addresses these problems by offering a simple and deterministic project structure where heavy data inspection and analysis are performed only once per unique dataset fingerprint.
The framework is suitable for:
- Individual ML practitioners
- Research prototyping
- Academic projects
- Small to medium ML teams
- Dataset registration and fingerprinting
- Persistent metadata storage
- Automated exploratory data analysis
- Experiment run management
- Model artifact registry
- Standardized prediction interface
- Evaluation and metric comparison
- Cleanup and pruning utilities
mlvern follows these principles:
- Deterministic dataset fingerprinting
- One-time heavy data inspection
- Minimal and explicit APIs
- Clear artifact organization
- Easy comparison between runs
- Simple prediction and evaluation helpers
Install the latest stable version from PyPI:
pip install mlvern