Skip to content
Latchkey

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

Run this faster and cheaper on Latchkey managed runners. Start free →