• Snowflake {{mayExportModel('jar') && !cantExportToSnowflake ? '' : '(Not available)'}}
  • Java {{mayExportModel('jar') ? '' : '(Not available)'}}
  • Python {{mayExportModel('python') ? '' : '(Not available)'}}
  • MLflow {{mayExportModel('python') ? '' : '(Not available)'}}

  • It is now possible to export a DSS Model trained in the Lab or imported from an MLflow model to a Databricks Registry.

    Both Workspace Registry and Unity Catalog are supported.

    Databricks {{mayExportModel('python') ? '' : '(Not available)'}}
  • PMML {{mayExportModel('pmml') ? '' : '(Not available)'}}
How does this export work ?
  1. The model will be exported to a persistent Snowflake function.
  2. It can be used by any Snowflake user in order to score records within Snowflake.


Not available: {{cantExportToSnowflake ? 'Export to Snowflake is not available in Visual Analyses. Deploy your model to the Flow first' : model.javaExportCompatibility.reason}}
How does this export work ?
  1. The model will be exported for use in Java code.
  2. It can be used in any Java program in order to score records, outside of Dataiku.

The model is exported as a Java class. You need to enter its fully-qualified name.
Not available: {{model.javaExportCompatibility.reason}}
How does this export work ?
  1. The model will be exported for use in Python code.
  2. It can be used in any Python program in order to score records, outside of Dataiku.

Not available: {{model.pythonCompatibility.reason}}
Since the model was originally imported from MLflow, you can choose to export the original model or a modified version of it.
How does this export work ?
  1. The model will be exported as the original MLflow Model. Some extra DSS files may appear in the zip but they won't have any incidence.
  2. The model can be imported in any MLflow-compatible system.

  1. The model will be exported as an MLflow Model using "dss" flavor of MLflow. This flavor uses the dataiku-scoring python package to load the model.
  2. The model can be imported in any MLflow-compatible system.

The generated MLflow model is also compatible with the "python_function" flavor of MLflow.

Not available: {{model.pythonCompatibility.reason}}
How does this export work ?
  1. The model will be exported to the MLflow format
  2. Then registered as an artifact of a new run of the specified experiment
  3. Finally added as a new version of a registered model

The generated MLflow model is compatible with the "python_function" flavor of MLflow.

Not available: {{model.pythonCompatibility.reason}}
{{uiState.progress || "Fetching experiment list..."}} A new experiment will be created. The exported model will be logged as an artefact of a run of this experiment.
{{uiState.progress || "Fetching model list..."}} A new registered model will be created if needed and the exported model will be added as a new version
How does this export work ?
  1. The model will be exported as a PMML file.
  2. The generated PMML file can be imported in any PMML-compatible scoring system.

Not available: {{model.pmmlCompatibility.reason}}
There are no export formats compatible with this model.