sce-data

name	sce-data
description	SCE dataset management, loading, and anonymization. Use when loading datasets, anonymizing sources, managing parquet files, choosing storage strategy, or validating checksums. Triggers include dataset, parquet, download, anonymize, checksums, Hugging Face.

SCE Data Handling

Quick Reference

Document	Purpose
references/DATASETS.md	Anonymized dataset map
references/ANONYMIZATION.md	Naming rules

When to Use This Skill

• You need to add or rename datasets
• You need to define storage strategy (repo vs remote)
• You need to document data access for CI

Procedure

1. Use anonymized names from DATASETS.md.
2. Store small datasets in repo as parquet.
3. Host large datasets remotely and provide a download script.
4. Verify checksums before use.

Gotchas

• Do not mention source brand names in public docs or configs.
• Do not commit large datasets to git history.
• rental_poland_long has development_floor fully missing; exclude it from numeric features or it will drop all rows.