Blame for START_HERE.md

# Upload ŊHAÏṢ–DQ II to Hugging Face

1. Extract the entire ZIP to a normal folder.
2. On Windows, double-click `UPLOAD_PUBLIC_TO_HF.bat` for a new public dataset, or `UPLOAD_TO_HF.bat` for a new private dataset.
3. Enter your repository ID, for example `YOUR_USERNAME/Nhais-DQ-II`.
4. Supply a Hugging Face token with write access when prompted. Input is hidden. The uploader also supports `HF_TOKEN` or existing Hugging Face authentication.

Python **3.10 or newer** is required. The launcher creates a local `.venv-hf` and installs the upload dependency. You do not need Git, a GPU, an API subscription, or to put your token in any file. Token settings: <https://huggingface.co/settings/tokens>.

The uploader checks the file manifest, creates a dataset repository if needed, and commits the unpacked files. It does not upload the outer ZIP, save the entered token, or delete remote files. Re-running updates matching file paths. Existing repository visibility is retained; the public launcher does not change a private repository's visibility.

If the launcher reports a missing Python installation, install Python 3.10+ with the Windows launcher or PATH option enabled, then run it again.

On macOS or Linux:

```sh
python3 -m venv .venv-hf
.venv-hf/bin/python -m pip install -r requirements-upload.txt
.venv-hf/bin/python upload_to_hf.py --repo-id YOUR_USERNAME/Nhais-DQ-II --public
```

To check everything offline before uploading:

```sh
python verify_release.py
python upload_to_hf.py --verify-only
```

The manuscript and executable research are under `research/`. The default displayed table has 24 experimental cases; the `generations` configuration has 192 generation records. Author credit is preserved exactly. No new license has been assigned.