ADR: Why Hugging Face Datasets?
Context: We need a place to store our Data Lake (Parquet files).
Options Considered: AWS S3, Google Cloud Storage, Local Filesystem, Hugging Face Hub.
Decision: Hugging Face Hub.
Rationale:
- Cloud storage (S3/GCS) costs money and requires setting up IAM roles, making reproduction harder for others.
- Hugging Face provides free hosting for open datasets, complete with version control and an API to download the data directly into Pandas/Polars.