ML-STAC Sample Specification 📜

The ML STAC Sample is the most important element in an ML-STAC implementation. An Sample constitutes the container for both tensors and metadata related to a single element within a specific dataset. Regardless of the dataset, all Samples share identical metadata fields. All the Samples must be stored in a safetensor format. For further details on why this format was chosen, please refer to the safetensor document.

In this directory 📁

Specification 📄: The ML-STAC Sample specification is in specification.md. It includes an overview and an explanation of the metadata and tensor fields.

Data Model 🧊: The dataclass.py contains the ML-STAC Sample datamodel definition. This can be used to generate JSON Schemas and validate ML-STAC Samples.

Schema 🧩: The schema.json file contains the ML-STAC Sample JSON Schema. This can be used to validate ML-STAC samples.

example.py 🐍: The example.py file contains an example of a valid ML-STAC Sample.