Reproducible Pipeline - CI/CD Glossary Definition
A reproducible pipeline yields identical outputs on rerun by pinning code, data versions, dependencies, and random seeds.
A reproducible pipeline produces identical outputs when rerun with the same inputs, pinning code, data versions, dependencies, and random seeds. It is essential for auditable data and ML work.
Reproducibility lets you rerun a result months later and get the same answer, which is a prerequisite for debugging and audit.
What to pin
Pin the code commit, the exact training data snapshot, the dependency lockfile, container image digests, and random seeds. Ephemeral runners help keep runs clean.
Related guides
Idempotent Pipeline - CI/CD Glossary DefinitionIdempotent Pipeline: An idempotent pipeline produces the same end state whether it runs once or many times on…
Experiment Tracking - CI/CD Glossary DefinitionExperiment Tracking: Experiment tracking records the parameters, code version, data version, metrics, and art…
MLOps - CI/CD Glossary DefinitionMLOps: MLOps applies CI/CD, testing, monitoring, and automation practices to the machine-learning lifecycle,…