How to Pull DVC-Tracked Data in GitHub Actions

dvc pull restores the exact data and model files pinned by the .dvc metadata committed to Git, using a configured remote and secret credentials.

Install DVC, configure the remote credentials from a secret, then run dvc pull to materialize the data version referenced by the current commit before training or testing.

Steps

• Install DVC with the remote extra (e.g. dvc[s3]).
• Set remote credentials from repository secrets.
• Run dvc pull to fetch the pinned data.

Workflow

jobs:
  data:
    runs-on: ubuntu-latest
    env:
      AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
      AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
    steps:
      - uses: actions/checkout@v4
      - run: pip install "dvc[s3]"
      - run: dvc pull -j 4

Gotchas

• Commit the .dvc files and dvc.lock so CI pulls the exact version you tested locally.
• Use -j to parallelize the pull for many small files.

Related guides