Faceberg: HuggingFace Datasets as Apache Iceberg Tables

Faceberg is a data-access utility created by kszucs in February 2026 that exposes HuggingFace datasets as Apache Iceberg tables. It provides a standardized interface for querying and managing the diverse collection of datasets hosted on the HuggingFace Hub using a high-performance table format.

Use Cases

• Querying HuggingFace text datasets using SQL engines via the Iceberg catalog
• Managing dataset versioning for ML models using Iceberg snapshots
• Applying schema evolution to HuggingFace tabular data using Iceberg metadata

Strengths

• Apache-2.0 license
• Apache Iceberg table format support
• Integration with HuggingFace Hub

Limitations

• No native data records (acts as a proxy)
• Dependent on HuggingFace API availability
• Requires Iceberg-compatible compute infrastructure

Provenance

Source

HuggingFace Hub

Collection Method

API-based exposure

Freshness

Last updated February 2026

Geography

Global