DataTerminology.comNEW
Also known as: DVC
Versioning datasets and models alongside code for reproducible machine learning workflows.
DESCRIPTION
Data Version Control (DVC) extends Git workflows to track data, models, and experiments. It stores large artifacts in external remotes and uses lightweight metafiles in Git. Reproducible pipelines define stage dependencies and enable incremental computation. Teams can compare experiments, share artifacts, and ensure consistent environments across machines. DVC integrates with CI/CD, cloud storage, and experiment dashboards to operationalize ML workflows.
ALTERNATIVE DEFINITIONS
DVC
DVC is an open-source version control system for machine learning projects.
Highlights open-source version control tailored to ML projects.
Iterative (GitHub)
Data Version Control | Git for data & models.
Positions DVC as Git-like workflows for data and models.
RELATED TERMS
Git Model Registry Artifact Store MLflow Experiment Tracking Reproduction Pipelines Lineage
COMPETING TERMS
lakeFS Pachyderm Git Large File Storage
RELATIONS
- Broader: Version Control
KNOWLEDGE AREAS
MLOps Data Engineering Platform & Tooling Data Management
TARGET AUDIENCE
Data Scientist MLOps Engineer AI Engineer Data Engineer
EXTERNAL SOURCES
- DVC - Official DVC introduction
- Iterative - DVC source code and docs
- Google Cloud - MLOps guide on data versioning
TOP QUESTIONS
Why use Data Version Control in ML projects?
It ensures reproducibility by tracking datasets, models, and pipelines alongside code, enabling consistent results across environments.
Category: Relevancy
How does DVC work with Git?
DVC stores pointers to large files in Git while pushing data to remote storage, keeping repositories lightweight.
Category: Mechanism
What types of storage can DVC use?
S3, GCS, Azure Blob, SSH, and local remotes are supported for artifact storage.
Category: Capabilities
How does DVC support pipelines?
DVC defines DAGs with stages and dependencies, enabling incremental runs and reproducible experiment execution.
Category: Concepts
How is DVC different from Git LFS?
DVC adds pipelines, metrics, and experiment tracking, whereas Git LFS focuses on storing large files in Git.
Category: Comparison
Can DVC integrate with CI/CD?
Yes. DVC pipelines and remotes fit into CI jobs to validate data, reproduce runs, and publish artifacts.
Category: Integration
What are common mistakes using DVC?
Committing large artifacts to Git, misconfiguring remotes, or not pinning data versions break reproducibility.
Category: Mistakes
Does DVC track experiment metrics?
Yes. It can log metrics and parameters for comparisons across runs and branches.
Category: Capabilities
What is a good starting point with DVC?
Initialize a project, configure a remote, add data with dvc add, and define a simple pipeline stage.
Category: HowTo