docs/UPLOAD_GUIDE.md

Upload EIROQ-PHI to Hugging Face

This folder is prepared as a Dataset repository containing a theoretical and computational research release. It includes machine-readable research tables, source code and the manuscript. It is not a trained model or a hosted application.

Suggested new repository name: EIROQ-PHI. Replace your-username below with your own Hugging Face account or an organization where you have write permission. No account or token is stored in this package.

Option 1: Browser upload

1. Sign in at https://huggingface.co and choose New Dataset at https://huggingface.co/new-dataset.
2. Select your namespace, enter EIROQ-PHI, and choose Private initially. Choose a license only if you have decided the release's actual license; this package makes no license grant.
3. In the new repository, open Files and versions → Add file → Upload files.
4. Upload the contents of the extracted release folder, including README.md, the machine-readable data directories, research/, docs/, tools/, the helpers and SHA256SUMS.txt. Preserve the folder hierarchy: README.md belongs at the repository root. Do not upload only this package's ZIP or the enclosing folder as one nested directory.
5. Commit the files, inspect the rendered Dataset card, and check that the manuscript and data links open. You can change the repository visibility in its settings when ready.

If your browser uploader does not preserve directory paths, use the helper below instead.

Option 2: Validated Python upload

Requires Python 3.10 or newer. The research calculations have separate dependencies in research/requirements-research.txt; uploading does not require installing them.

From a terminal in the extracted release folder, install the upload client once:

python -m pip install --upgrade huggingface_hub

On Windows, use py -3 in place of python if needed:

py -3 -m pip install --upgrade huggingface_hub

Check the package without signing in or connecting to the network:

python upload_to_hf.py --repo-id your-username/EIROQ-PHI --dry-run

Sign in using hf auth login, or let the helper securely prompt through the Hugging Face client when you start the upload. Use an access token with write permission for your chosen namespace. Never put tokens in the scripts, command arguments, Dataset card or uploaded files.

Create a new private Dataset and upload the validated release:

python upload_to_hf.py --repo-id your-username/EIROQ-PHI

To create it publicly, explicitly add --public:

python upload_to_hf.py --repo-id your-username/EIROQ-PHI --public

Windows convenience commands:

.\UPLOAD_WINDOWS.bat --repo-id your-username/EIROQ-PHI --dry-run
.\UPLOAD_WINDOWS.bat --repo-id your-username/EIROQ-PHI

macOS/Linux convenience commands:

sh UPLOAD_UNIX.sh --repo-id your-username/EIROQ-PHI --dry-run
sh UPLOAD_UNIX.sh --repo-id your-username/EIROQ-PHI

Double-clicking UPLOAD_WINDOWS.bat also starts the helper and prompts for the repository id. The default visibility is private.

What the helper checks and changes

The helper runs tools/validate_release.py, verifies the root checksum manifest, stages only manifested files, and uploads that explicit allowlist plus SHA256SUMS.txt. It excludes any unrelated files added beside the release, including caches. A dry run performs local checks only and needs no Hugging Face package or authentication.

An upload creates a new Dataset repository using exist_ok=False. It deliberately stops if that repository already exists and does not overwrite existing work. Choose another new name if necessary. If repository creation succeeds but a later upload fails, the helper reports that incomplete state; inspect the new repository in the browser before taking further action.

Keep the extracted release unchanged until upload. If checksum validation fails, re-extract a clean copy. After you intentionally modify release files, rebuild the manifest and run the validator before uploading.

Scientific and licensing status

Keep the Dataset card's assumptions, synthetic-data labels and limitations intact. The archive provides theoretical results and executed computational checks; it is not laboratory evidence, a demonstrated device, or an established field-wide breakthrough. Historical novelty and priority remain unconfirmed.

The original release does not specify a license. Uploading to Hugging Face does not by itself grant a reuse license. No LICENSE file or open-license metadata has been invented for this preparation.