DVC (Data Version Control) is an open-source command-line tool that versions datasets and models alongside your code in Git, so machine learning projects can be reproduced.
In plain words
Git is great for code but poor at storing huge data files. DVC keeps the big files, such as datasets and trained models, in cloud storage and puts small pointer files into Git that record which version you used. It can also describe your pipeline, the steps that turn data into a model, and rerun only the steps affected by a change. Experiments are tracked in your local repository without a server.
A quick example
Imagine a team training an image classifier on thousands of photos. They add the photo folder to DVC and push the files to a cloud bucket. Months later, anyone can check out an old Git commit, pull the matching photos with DVC, and rebuild exactly the same model.
Where you'll meet it
- Versioning datasets and trained models next to code.
- Reproducible machine learning pipelines and experiments.
Related
Sources
- Docstreeverse/dvc, DVC (GitHub) · read 29 Sept 2026
- DocsDVC documentation, DVC · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.