Skip to main content

DDS Adoption Guide

Use DDS when collecting, transforming, preserving, or releasing a dataset.

Steps​

  1. Create DATASET.manifest.toml.
  2. Record source provenance.
  3. Record license and usage limits.
  4. Define validation procedure.
  5. Record hashes for preserved artifacts.
  6. Document splits and known limitations.
  7. Identify dataset class and link derived datasets to sources.
  8. Record sensitive-data, license, or usage constraints before release.

SFDS Relationship​

Use SFDS to maintain DDS as a standard suite. Use DDS to govern dataset provenance, license status, validation evidence, hashes, splits, and usage limits.