Model Validation - CI/CD Glossary Definition
Model validation checks a trained model against quality thresholds on held-out data and blocks promotion when it fails.
Model validation evaluates a trained model on held-out data against quality thresholds (accuracy, precision, recall, calibration) before it can be promoted. Failing thresholds block deployment.
Validation is the CI gate for models: it turns evaluation metrics into a pass or fail decision in the deployment pipeline.
Gate on metrics
A validation stage asserts the new model beats a baseline and meets minimum metrics on a test set (test-set-gl36). If not, the pipeline stops.
Related guides
Test Set - CI/CD Glossary DefinitionTest Set: A test set is data held out and touched only once, at the end, to estimate how a finished model wil…
Model Training - CI/CD Glossary DefinitionModel Training: Model training is the process of fitting a machine-learning model to training data by optimiz…
Model Deployment - CI/CD Glossary DefinitionModel Deployment: Model deployment is the act of releasing a validated model to production so it can serve pr…