Model Training - CI/CD Glossary Definition

Model training fits a model to training data by optimizing parameters, producing a versioned model artifact.

Model training is the process of fitting a machine-learning model to training data by optimizing its parameters against a loss function. The output is a versioned model artifact.

Training is often run as an automated pipeline stage triggered by new data or code, with results logged for reproducibility.

As a pipeline stage

Automated training reads a pinned dataset version, runs on a runner or GPU pool, and writes the resulting model plus metrics to a registry for later validation.

Related guides