- user-authored Python notebook and validation utilities;
- kernel metadata and data-mount smoke tests;
- compact run/provenance receipts that contain no competition samples;
- links and checks for obtaining the original data from Kaggle.
raw/,processed/,submissions/,runs/, andlogs/;- archive files, model weights, and generated prediction artifacts;
- virtual environments, bytecode, credentials, and machine-local settings.
After gaining access through the official competition page, acquire data into a private local directory, set the paths expected by the notebook, and run the included validation commands. Never use this public repository as evidence that a reader is permitted to obtain, redistribute, or train on the original competition data.