Install and verify
Download
- Full data package (Hugging Face): https://huggingface.co/datasets/Zafrens/zscreen-pilot
- Code and documentation (GitHub): https://github.com/Zafrens/zscreen-pilot
- Archived release (Zenodo, resolves to the latest version): https://doi.org/10.5281/zenodo.22003566
Use the full package folder or matching full archive. Open a terminal in the extracted package root. Python 3.11 or later is supported:
python -m pip install -e .
python verify.py --full
The verifier checks every distributed file against the root file_manifest.csv and MANIFEST.json, then checks data shapes, axes, IDs and current package objects. These two inventory files are excluded from their own checksum list: total files equal manifested files plus two. Recognized local virtual environments, caches and OS metadata are excluded; unlisted payloads and altered distributed files fail.
Reference model
For the strict fixed-prediction check, use the verified CPU build:
python -m pip install torch==2.5.1 --index-url https://download.pytorch.org/whl/cpu
python -m pip install -e ".[model]"
python models/predict.py --check-golden
The diagnostic applies its reference software thread count and restores the caller's setting. It does not require 12 hardware cores. Model guide.
To reproduce the complete dependency snapshot, use Python 3.11.14, install the CPU PyTorch build above, then run:
python -m pip install -r environment.lock
python -m pip install -e . --no-deps
For notebooks with supported dependency ranges, install python -m pip install -e ".[notebooks]".
Archive checksums
If supplied with the archive, SHA256SUMS.txt uses LF line endings. Run sha256sum -c SHA256SUMS.txt on Linux or shasum -a 256 -c SHA256SUMS.txt on macOS from the archive directory. On Windows compare Get-FileHash -Algorithm SHA256 with the list.
A code/documentation review subset omits large inputs; use the full download for execution. Start Here · Data dictionary · Data access.