How to develop flwr-datasets¶

Flower Datasets uses uv for development and CI. The commands presented in this guide assume you have cloned the Flower Github repository.

Setup¶

cd datasets
uv sync --all-extras

Tip

Use uv sync --frozen --all-extras to ensure uv.lock is not modified.

Run checks (formatting + unit tests)¶

cd datasets
uv run ./dev/test.sh

Format¶

cd datasets
uv run ./dev/format.sh

Build docs¶

cd datasets
uv run ./dev/build-flwr-datasets-docs.sh

Run E2E tests¶

cd datasets/e2e/pytorch
uv sync --frozen
uv run python -m unittest discover -p "*_test.py"

Repeat for datasets/e2e/scikit-learn and datasets/e2e/tensorflow.

Dependency management (no `uv pip`)¶

cd datasets

# Add a runtime dependency
uv add <package>

# Add a dev dependency
uv add --dev <package>

# Add a dependency to an extra (e.g. "vision")
uv add --optional vision <package>

# Update lockfile (commit the result)
uv lock