Stable Datasets

Stable Datasets is a remote-accessible data repository released by the galilai-group and last updated in March 2026. It is designed for direct integration with the Hugging Face Datasets library and PyTorch workflows, facilitating remote data streaming. The repository provides a standardized interface for galilai-group's data assets under the Apache-2.0 license.

Use Cases

• Implementation of remote data streaming using the 'remote download support' feature
• Integration into PyTorch-based machine learning pipelines via the huggingface-datasets interface
• Benchmarking galilai-group research outputs using their standardized data loaders

Strengths

• Apache-2.0 permissive license allows for commercial and private use
• Native support for remote download functionality via Hugging Face Datasets
• Direct compatibility with PyTorch training pipelines

Limitations

• Lack of documented column schemas or record counts in the primary metadata
• Absence of sample data prevents pre-download inspection of data quality

Provenance

Source

galilai-group GitHub repository

Freshness

Last updated March 13, 2026.