In addition to tracking model parameters, trained models themselves can also be tracked across the model development lifecycle via a version control system. Versioning models allows data science and machine learning teams to more readily reproduce experiments and provides for greater consistency between prototyping and production environments. Models may be versioned by checking model code, serialized model objects (e.g. a Python .pkl file), a model’s Docker container, or other model artifacts into a Git repository, ideally automatically at appropriate points in the modeling workflow. Most modern ML platform offerings support versioning models.