dvc/YouTube/Tutorial Playlist/01 Versioning Data with DVC (Hands-On Tutorial!)

• Versioning Data with DVC (Hands-On Tutorial!)

  • The opening video in dvc/YouTube/Tutorial Playlist introduces how DVC extends Git projects to version large data and models stored outside the Git repository.

  • Notes

    • Initialize Git and DVC, obtain a sample dataset with dvc get, then use dvc add to create a small .dvc file that Git can track while ignoring the dataset itself.
    • Configure a DVC remote, push the dataset to it, and pull a copy after removing the local data and cache.
    • After changing the dataset, run dvc add again, commit the changed .dvc file, and push the new data. Checking out an earlier Git revision of that .dvc file and running dvc checkout restores the corresponding data version.
    • Git records the version history of the pointer files; DVC stores and retrieves the larger data. The same workflow also applies to models and other files.

  • Readwise highlights

    • dvc get retrieves a dataset from a DVC data registry.
    • dvc remote add -d configures a default storage remote; the tutorial demonstrates Google Drive.
    • Git tracks the .dvc metadata while DVC manages the large file in remote storage.