This page covers how to manage artifacts and build prediction services on models.

Artifacts & Models

Artifact is a superset of a dataset and can be in different formats such as zip file, folder, xls, parquet, and so on. You can upload the artifact file to the canvas flow and run recipes on this artifact. However, the recipes or templates should be configured to use artifacts as input in a flow. Running this recipe also generates an artifact output. On the other hand, the trained model is an output generated after running a model builder transform in the machine learning pipeline. All the models generated in projects under a particular tenant are saved in the model catalog. You can later use the models as pipeline input to make predictions on live data.

Creating an artifact

Use this procedure to create an artifact and add multiple files to an artifact.

To create an artifact:

1. Click the menu icon and select Artifacts & Models.

The Artifacts tab is displayed.

2. Click the plus icon. If the page does not have any artifact, you can view +Create Artifact option in the workspace. The Create Artifact window appears.

3. Specify the artifact name in the Artifact Name field.

4. Click BROWSE FILE to browse and upload the file to this artifact folder from your local system. You can view the artifact file added with the file name, file size and file type after adding.

5. Click Create Artifact. The artifact gets created and you can view this on the Artifacts tab

Important

You cannot delete the artifacts that are used in the canvas flow.

6. If you want to add another file to an existing artifact, click on the artifact. This artifact page is displayed showing the existing files associated with this artifact in the Files section.

7. Click the plus icon.

This opens the Add files window.

8. Click BROWSE FILE to browse and upload another file to this artifact.

9. Click Add Files to add files to the artifact. You can follow the same procedure to upload multiple files to this artifact.

Managing artifacts

Use this procedure to manage artifacts in a tenant.

To manage artifacts:

1. Select Artifacts & Models from the left navigation menu. The Artifacts tab is displayed where you can view the list of all artifacts in a tenant.

2. Review this information:

Name:

The name of the artifact.

Created:

The date on which the artifact is uploaded.

Source:

The source from where the artifact has been uploaded. Possible values: * Manual addition * Name of the transform(Project name) where artifact is passed as an input

Note

When you rest the pointer on the artifact name, an ellipses icon is displayed. Select delete to remove the artifact from the list permanently.

3. Click on a particular artifact name whose details you would like to view. The files tab is displayed.

4. Review this information:

File:

The name of the artifact file.

File size:

The file size of the artifact file.

File Type:

The type of artifact file.

Actions:

You can perform these actions:

• download the artifact file, using the download icon.

• delete the artifact file, using the delete icon.

• add a new file to a specific artifact, using the +ADD FILE option.

• Use the table settings icon to reorder the columns or select and deselect the columns you want to view in the table.

Note

You can use the search option to find a specific artifact from the list.

Viewing models

Use this procedure to view list of trained models generated in all the projects in a tenant. You can use the search option to look for model details you want.

To view models:

1. Select Artifacts & Models from the left navigation menu. The Artifacts tab is displayed.

2. Click the Models tab to view the list of models generated after running the machine learning pipeline in different projects.

3. Review this information:

Name:

The name of the machine learning training model.

Created:

The time stamp at which the model was generated.

Source:

The project in which this training model is generated.

Prediction service:

Click the +Add to create a prediction service. To add a prediction service, see Creating a prediction service.

Deleting a model

Use this procedure to delete a model.

To delete a model:

1. Click the Models tab to see the list.

2. Click the ellipses icon corresponding to the model you want to delete and then select DELETE.

Note

You cannot delete the model that is created in a project, but can delete the models that are added manually.

Prediction services

Prediction Service allows you to send real-time data to a model and receive predictions immediately. You can create an endpoint of the model that is exposed as an API and upload the test dataset. This API that has the model that makes predictions on the uploaded data.

Creating and testing a prediction service

Use this procedure to create a prediction service or an endpoint of the model. This service can only be created for models generated after running the data pipeline in a project (or) on models that are manually added. After generating the model as an API, you can test this model on the uploaded dataset to make predictions.

1. Select Artifacts & Models from the left navigation menu. The Artifacts tab is displayed where you can view the list of all artifacts in a tenant.

2. Click the Models tab to see the list of models created in this tenant.

3. Click +Add in the Prediction Service column and corresponding to the model whose API you want to create. This displays the Prediction Service tab.

4. Specify this information in the Details section:

Name:

The name of the prediction service.

Description:

The description of the prediction service.

Environment:

The environment in which you want to test the prediction service.

Pre-process & Post-process:

If needed, you can add pre-processing and post-processing steps with the integrated code editor.

5. Specify the Configuration Options:

Timeout:

The duration (in minutes) after which the incoming request should time out.

Concurrency:

The number of parallel requests you can send. The acceptable number of requests you can send at a time ranges from 5 to 100.

6. Enable the Save Logs option to view a detailed record of activities that occurred in the prediction service. Disable this option to stop tracking logs.

7. Click Save to create the endpoint for the model.

This generates a unique endpoint for the model and the CURL command.

8. Select the file formats. Possible values:

• JSON

• CSV File

• Canvas Datasets

9. Click Browse to upload the file either in the CSV or JSON format based the selected file format.

10. Click Test to check the prediction results.

After testing the prediction service on an uploaded dataset, you will see two options: Download as CSV and Add to Canvas. You can either download the CSV file to view the output using the ‘Download as CSV’ option, or add the dataset to the canvas using the ‘Add to Canvas’ option

Note

You can check the logs from the last 30 days by clicking the Logs option. This will show information about the types of queries executed, the number of successful queries, and the number of failed queries. However, you can only view the logs if the Save Logs toggle is enabled.