RIME Registries

RIME provides registries to keep track of the models, datasets, and predictions files that have been uploaded to a RIME instance. You can register these data objects from the RIME SDK or in the RIME web UI during project configuration.

Models are typically Python files in either py or pkl formats. Reference or evaluation datasets and prediction sets can be provided in several formats, most commonly CSV.

Any of these data objects can be directly uploaded to the RIME instance using the SDK or sourced from an integration. Register the data object after upload.

Once a model, dataset, or prediction set has been registered with a RIME instance, that data object is available for selection in the appropriate drop-down selectors during project configuration for either a Stress test or a Continuous test.

Registering a model, dataset, or prediction set with a RIME instance using the SDK

This procedure assumes the RIME SDK is installed and available for use and that the relevant data objects have been uploaded to the RIME instance. See the Python SDK documentation for full details on the SDK calls used in this procedure.

  1. Choose an existing project or create a new project.

    Option

    Action

    Existing project

    Use the client.get_project call with the ID of the project.

    New project

    Use the client.create_project call to create the new project.

  2. Use the correct register call based on the type of object being registered.

    Data object type

    SDK call

    Model

    project.register_model

    Dataset

    project.register_dataset

    Prediction set

    project.register_predictions

The registered data objects are now available from the UI during project configuration.