DSLab Runner

The DS Lab Runner component is used to manage and execute data science experiments that were created in the Data Science Lab (DSLab) module and imported into a pipeline. It supports running both trained models and exported scripts as part of a data pipeline workflow.

The DS Lab Runner processes input data from one event and sends the transformed output to another event. It can operate in batch or real-time mode, depending on the selected configuration.

Requirements

Before using the DS Lab Runner, ensure that:

  • You have created and saved a model or script in the DS Lab module.

  • Input and output events are available in the pipeline canvas.

  • Compute resources (CPU/GPU) are configured for model execution.

Using the DS Lab Runner in a Pipeline Workflow

  1. Drag the DS Lab Runner component to the Pipeline Workflow canvas.

  2. Create two events (input and output) and place them on the canvas.

    • Input data can originate from ingestion components, readers, shared events, or scripts from the DS Lab module.

  3. Connect the input eventDS Lab Runneroutput event.

  4. Select the DS Lab Runner component to open its configuration tabs.

  5. Configure the fields in the Basic Information and Meta Information tabs.

  6. Save the component.

Once configured, the DS Lab Runner reads input data, runs the model or script, and writes the processed data (including predicted columns, if applicable) to the output event.

Configuration

All configurations are grouped into the following sections:

  • Basic Information

  • Meta Information

  • Resource Configuration

Basic Information Tab

The Basic Information tab defines general properties and execution settings.

Field
Description
Required

Invocation Type

Select execution mode: Batch or Real-Time.

Yes

Grace Period (sec)

Appears only for Batch mode. Defines the time before the component shuts down gracefully.

Conditional

Deployment Type

Pre-selected field showing the deployment type of the component.

Yes

Batch Size

Maximum number of records processed per cycle. Minimum: 10.

Yes

Failover Event

Select an event to handle failover scenarios.

Optional

Container Image Version

Displays the Docker image version used. Pre-selected.

Yes

Description

Optional description of the component.

No

Meta Information Tab

The Meta Information tab determines the execution type:

  • Model Runner (run a registered model from DSLab).

  • Script Runner (run an exported script from DSLab).

DS Lab Runner as Model Runner

Use this mode to execute a model created in the DS Lab module.

Field
Description
Required

Project Name

Name of the project containing the model.

Yes

Model Name

Name of the saved model in the project.

Yes

DS Lab Runner as Script Runner

Use this mode to execute a Python script exported from the DSLab module.

Field
Description
Required

Function Input Type

Select the input type: DataFrame or List of Dictionary.

Yes

Project Name

Name of the project containing the script.

Yes

Script Name

Name of the exported script from DSLab notebooks. Must be written inside a function.

Yes

External Library

Comma-separated list of external libraries used in the script.

Optional

Start Function

The function name defined in the script.

Yes

Script

Displays the exported script content.

Yes

Input Data

Key-value pairs for function parameters.

Optional

For details on exporting a script, see: Exporting a Script from DSLab.

Saving the DS Lab Runner

  • After completing the configuration, click Save Component (Storage icon).

  • A success message confirms that the component has been saved.

Example Workflow

  1. Ingest customer transaction data through a Reader component.

  2. Pass the data to a DS Lab Runner (Model Runner) that applies a fraud detection model.

  3. Write the enriched dataset (with predicted fraud risk scores) to an output Event.