Import a Model

External models can be imported into the Data Science Lab and experimented inside the Notebooks using the Import Model functionality.

The Import Model functionality in the Data Science Lab allows you to bring in external models and experiment with them inside notebooks. Imported models are identified as External models and can be registered to the Data Pipeline module for inference.

Important:

Only native prediction functionality works for external models.
Supported file extensions: .pkl, .h5, .pth, .pt.

Import an External Model

Follow these steps to import a model into the Data Science Lab:

Navigate to the Data Science Lab module.
In the left panel, click Models.
- The Models page opens, listing available models.
Click Import Model.
Upload the model file from your system.
A notification message confirms the import.
The imported model is added to the Models list and is:
- Marked with the letter E for easy identification.
- Labeled as Ext under model type.
Select the imported model to view details in the right-side panel.

Exporting an Imported Model to the Data Pipeline

Imported models can be registered and used within the Data Pipeline through the Script Runner component.

Prepare and Register the Model

Open a Data Science Notebook (.ipynb file) from an activated project.
Write a wrapper function that includes:
- Input data
- Imported model
- Predict function
- Output dataset with predictions
Load the imported model into a notebook cell.
Reference the loaded model in your inference script.
Run the notebook cell to generate predictions.
From the ellipsis menu of the model, click Register.
In the Register Model dialog, confirm the action by clicking Yes.
A notification confirms successful registration.

Export the Script to Pipeline

From the notebook toolbar, select Export.
Save the notebook when prompted.
The Export to Pipeline window opens.
Select a specific script or choose Select All.
Click Next.
Click Validate to check the script.
- A notification confirms script validity.
Click Export to Pipeline.
A notification confirms successful export.

Accessing the Exported Model in the Data Pipeline

Navigate to the Data Pipeline Workflow Editor.
Drag the DS Lab Runner component onto the canvas.
Configure the Basic Information.
Open the Meta Information tab and configure:
- Execution Type: Select Script Runner.
- Function Input Type: Choose the appropriate input type.
- Project Name: Select the project where the model resides.
- Script Name: Select the script name (matches the imported model name).
- External Library: Provide details if required.
- Start Function: Select the entry point function.
Connect the DS Lab Runner to an Input Event.
Run the pipeline.
- Model predictions appear in the Preview tab of the connected event.

Key Notes

Imported models are accessible through the Script Runner in Data Pipeline.
Use Model Runner execution type for native Data Science models.
External models are always labeled with E and Ext for clarity.
Ensure correct environment and library dependencies are configured in notebooks.