Knowledge Maxxing← Back to topic library
ML MaxxingAdvanced75 sec
Data versioning
Data versioning tracks changes to a dataset over time, so a specific model can always be linked back to the exact data it was trained or evaluated on.
Prerequisite: Experiment tracking
What matters
Without versioning, a dataset that later changes or grows makes it impossible to reproduce a past experiment's exact results or to trust a comparison across time.
Example
A data-versioning system might record that model v3 was trained on dataset snapshot v7, so anyone can later reproduce that exact training run.
Common misconception
Data versioning is not the same as simply keeping a backup copy of a dataset; it specifically tracks which version was used for which experiment or model.
Recall question
Why does an unversioned, changing dataset make past results hard to reproduce?
Keywords and related concepts
data versioningreproducibilitydataset trackingexperiment tracking