Datasets:

PureOne
/

universal-programmable-matter-voxels

The ZIP contains the entire standalone research repository, papers, source code, results, machine-readable records and publisher. Target: PureOne/universal-programmable-matter-voxels, a public dataset repository containing the research and its synthetic data. No model weights or hosted compute are needed.

Windows: three steps

Extract the ZIP completely into a writable folder. Do not run the BAT from inside the ZIP preview.
Double-click PUBLISH_TO_HUGGINGFACE.bat. Python 3.10+ is required; Python 3.12 is recommended. The script installs the small upload dependency in a local virtual environment.
Paste your PureOne write token into the hidden prompt and press Enter. Nothing appears while you type or paste. The publisher verifies your account, uploads the complete release, creates tag v1.0.1, downloads all release files anonymously to verify them, and opens the public page.

The token must permit creating and writing the target repository. Create or inspect tokens at Hugging Face token settings. A fine-grained token limited to an already-existing repository may require you to create that public dataset repository first. An empty public repository is supported.

The publisher keeps the token in process memory; it does not save it in the BAT, ZIP, repository or publication receipt. Do not paste your token into any project file. An existing HF_TOKEN environment variable can be used instead of the hidden prompt.

This ZIP is prepared for publication. A successful run is the event that publishes it; preparation alone does not mean the repository is live. Look for Published and anonymously verified and publication_receipt.json, which records the actual commit, tag and verified file count without credentials.

What will be public

Only files listed in SHA256SUMS.txt, plus that checksum file itself, are staged and uploaded. That includes the complete research and the publisher source. Local virtual environments, extra personal files and the receipt are excluded by this explicit file list. New files you add are not uploaded automatically. Existing remote files are never deleted.

The destination account is checked against PureOne. A private, gated or unrelated existing repository is refused. The publisher does not change another repository's visibility, force-push or move an existing version tag. If interrupted, rerun it: the same release can be verified again without creating a different tag.

Command-line alternatives

Offline verification, no token or package installation:

python publisher/publish.py --dry-run

Publish from Windows, Linux or macOS:

python -m pip install -r publisher/requirements.txt
python publisher/publish.py --publish --open

Read README.md for the research, docs/EXPERT_REVIEW.md for expert review and AGENTS.md for AI-agent use. The paper PDFs remain scientific version 1.0.0; the public distribution is version 1.0.1.

Troubleshooting

Python missing: install Python from python.org, including the launcher, then rerun.
Checksum mismatch: extract a fresh ZIP. Do not edit the token into a file or modify the release before publishing it.
401 or 403: check the token account, expiry and repository write scope. The connected read-only account in ChatGPT does not grant the local script write access.
Network or tag failure after upload: rerun. The script checks for an already-published identical version and then verifies public bytes.
Existing different tag/project: use the intended empty public repository or prepare a new versioned release. The uploader deliberately does not overwrite unrelated content.
Dataset viewer still building: the files and README can be public before the optional data preview finishes. The success receipt verifies file access, not viewer indexing.

Publishing API behavior was checked against repository management, upload guidance, dataset-card metadata and dataset configuration. SDK calls were validated locally; the Windows launcher and actual authenticated upload require your Windows environment and write token.