TroveGrid
Back to all tools
DVC logo

DVC

Git for ML — version datasets, models, and pipelines alongside your code.

0Open Source
AI Data MLOps
Official Website View on GitHub
Share:XLinkedInWhatsApp

Overview

DVC (Data Version Control) is an open-source tool for versioning ML datasets, models, and pipelines using Git-like commands, enabling reproducible ML experiments.

Key Features

  • Dataset and model versioning with Git-like commands
  • Remote storage support: S3, GCS, Azure, SSH, HDFS
  • DVC Pipelines for reproducible ML workflows
  • Experiment tracking with dvc exp run
  • Data registry for sharing datasets across teams
  • CI/CD integration for automated ML pipelines

Real-World Workflows

Track and reproduce ML experiments

  1. 1Initialize DVC in your Git repo: dvc init
  2. 2Add dataset to DVC tracking: dvc add data/train.csv
  3. 3Define pipeline stages in dvc.yaml
  4. 4Run experiments: dvc exp run --set-param lr=0.01
  5. 5Compare experiments: dvc exp show

Getting Started

pip install dvc

# Initialize in a git repo
git init && dvc init

# Track a dataset
dvc add data/dataset.csv
git add data/.gitignore data/dataset.csv.dvc
git commit -m 'track dataset with DVC'

# Push data to remote storage
dvc remote add -d myremote s3://mybucket/data
dvc push

Compare Alternatives

See how DVC stacks up against similar tools.

VSmlflowVSwandbVSneptuneVSlakefs