Skip to content

Mirror Locations

Jump to bottom
puma-community-bot edited this page May 24, 2026 · 1 revision

Mirror Locations

PUMA Community submissions can optionally be mirrored to external archives for long-term accessibility, citation, and easy programmatic discovery. The canonical source is always this GitHub repository; mirrors are downstream copies refreshed automatically by GitHub Actions when the corresponding secrets are configured.

Planned mirrors

  • Hugging Face Datasets at pumacp/puma-community-submissions. Each accepted submission becomes a row in a Parquet-backed dataset. Use case: programmatic discovery, filtering, and analytical queries via the datasets library.
  • Zenodo community pumacp. Monthly DOI-backed snapshots of the full archive, suitable for academic citation.
  • Kaggle dataset at pumacp/puma-community-submissions. Mirror in the Kaggle dataset catalog so that Kaggle Notebooks can attach the archive directly.

Status

Mirrors are infrastructure-ready but not yet populated. Each mirror has its own GitHub Actions workflow under .github/workflows/ (mirror-huggingface.yml, mirror-zenodo.yml, mirror-kaggle.yml), each gated on a repository secret that the project maintainers have not yet provided. Until activation, the GitHub repository remains the canonical and only source.

How to query the mirrors once active

Once Hugging Face is populated:

from datasets import load_dataset
ds = load_dataset("pumacp/puma-community-submissions", split="train")

Once Kaggle is populated:

kaggle datasets download pumacp/puma-community-submissions

For Zenodo, browse the community at https://zenodo.org/communities/pumacp (once activated) and download the most recent snapshot directly from there. Each snapshot has a stable DOI suitable for citation.

Clone this wiki locally