Verified project record

treeverse/dvc

DVC provides Git-like version control for data and models, lightweight ML pipelines, and local experiment tracking without requiring servers. It addresses the challenge of versioning data, models, and ML experiments while ensuring reproducibility.

Project overview

By storing data artifacts in an external cache while keeping version metadata in Git, DVC brings familiar version-control ergonomics to machine learning workflows that Git-LFS and manual experiment logs cannot adequately serve.

Project type: Model Development · Data Processing
Deployment: Refer to project documentation
License: Apache-2.0

Best for

Developers who need Git-like version control for data and models without deploying a server.
Data teams and AI engineers who need lightweight ML pipelines that run only steps impacted by changes.
Machine learning practitioners who want to track and compare experiments locally in a Git repo without servers.

Key capabilities

Stores data and model artifacts in a cache outside of Git, keeping version info in the Git repo and supporting remote storage platforms.
Connects code and data together in computational graphs to specify all steps required to produce a model, running only the steps impacted by changes.
Tracks, filters, and compares ML experiments locally in a Git repo without needing servers.

Limitations and risks

Coding is required to use DVC, as it involves working with Python code and parameters.
Using remote storage platforms such as S3, Azure, or Google Cloud, or on-premise network storage, is optional for sharing artifacts.

Getting started

Installation difficulty is straightforward with multiple options available including pip, snap, brew, conda, and choco. After installation, run `dvc add` to track data.

Evidence and sources

GitHub project description: 🦉 Data Versioning and ML Experiments
README: **Data Version Control** or **DVC** is a command line tool and `VS Code Extension`_ to help you develop reproducible machine learning projects: #. **Version** your data and models…
README: To share and back up the *data cache*, DVC supports multiple remote storage platforms - any cloud (S3, Azure, Google Cloud, etc.) or on-premise network storage (via SSH, for examp…
README: The closest *analogies* to describe the main DVC features are these: #. **Git for data**: Store and share data artifacts (like Git-LFS but without a server) and models, connecting…
Release: 3.67.1

Reviewed at: 2026-07-17T14:03:38Z