ecg-k-fold 0.1.3

ECG k-Fold utilities

ECG-k-Fold

Utilities for creating reproducible k-fold splits for open-access ECG datasets (FAIR compliant), packaged as a Python library with a CLI and CI/CD. Includes an option to push datasets to the Hugging Face Hub.

Install

pip install ecg-k-fold

For development:

git clone <this-repo>
cd ECG-k-Fold
make init

Or manually:

git clone <this-repo>
cd ECG-k-Fold
# Install uv if not already installed: curl -LsSf https://astral.sh/uv/install.sh | sh
uv pip install -e .[dev]
uv run pre-commit install

CLI

Push a local dataset directory to the Hugging Face Hub. Requires an access token with write permissions.

export HF_TOKEN=hf_************************
ecg-k-fold push-dataset ./path/to/dataset \
  --repo-id org_or_user/dataset-name \
  --private \
  --commit-message "Add v1"

Flags:

• --repo-id: target dataset repo, e.g. your-org/ecg-dataset
• --private/--public: repository visibility (default private)
• --commit-message: commit title for the upload
• --token: override HF_TOKEN env var

Dataset contents are uploaded as-is; common temp files are ignored.

Python API

from pathlib import Path
from ecg_k_fold.datasets.push_hf import push_local_dataset_to_hub

push_local_dataset_to_hub(
    local_dir=Path("./path/to/dataset"),
    repo_id="your-org/ecg-dataset",
    private=True,
    commit_message="Initial upload",
    hf_token="hf_...",
)

CI/CD

• Lint, type-check, and tests run on PRs and pushes to develop and main (.github/workflows/ci.yml).
• Publishing to PyPI happens on pushing a tag like vX.Y.Z (.github/workflows/publish.yml). Uses PyPI Trusted Publishers, no PyPI token needed. See https://docs.pypi.org/trusted-publishers/
• A manual workflow can push datasets to Hugging Face (.github/workflows/hf-dataset-push.yml). Set HF_TOKEN in repo secrets.

Gitflow

• Default branches: main (stable), develop (integration).
• Create feature branches off develop (feature/<name>), open PRs into develop.
• Release by creating a tag vX.Y.Z on main to trigger PyPI publish.

See CONTRIBUTING.md for details.

License

Apache-2.0. See LICENSE.