Versions represent a specific version of the underlying synthetic training data that belongs to a dataset. Each individual version and is a large collection of sample pairs (input, output) that are in the form of the schema, spread across the knowledge graph, in the context of the task.

The first version of every dataset is designed in the dataset designer but subsequent versions are created by editing prior versions.

Each dataset can contain any number of versions, and each of these versions can have a model trained on its data.

Last updated