Deploying the trained Machine Learning model

Deploying trained Machine Learning models closer to the edge network, specifically on BMC Helix Edge nodes, is a strategic approach to enable real-time inferring on IoT data. By bringing Machine Learning to the edge, you can achieve the following by handling data generated by IoT devices:

• Make decisions locally
• Reduce latency
• Increase the efficiency

The deployment process involves optimizing the model's placement in the BMC Helix Edge node infrastructure for integration with the edge network for immediate and responsive IoT data streams. In this section, you explore the deployment process of a Machine Learning trained model, which involves: 

• Selecting a trained model
• Specifying deployment details
• Ensuring the deployed model operates in the environment

With deployment models, you can leverage predictive capabilities for real-time decision-making and automation. You also learn to remove a trained model from a node or a device.  

To deploy a trained Machine Learning model

The following video displays how to deploy a trained Machine Learning model:

Make sure that you have a trained Machine Learning model. To train a model, go to Training-a-machine-learning-model.

1. On the BMC Helix Edge page, select Machine Learning.
2. Identify the specific trained Machine Learning model to deploy on a node in the list of trained models.
3. Clickto expand the model console.
4. Identify the deployment version in the expanded trained Machine Learning model console and select Action > Deploy. 
   BMC Helix Edge displays the Deploy Model panel with trained models related to the deployment.

5. Review the following deployment model settings:

   Column heading	Description
   Node	Displays the node or location where the model is deployed.
   Node ID	Displays the unique identifier associated with the deployment node.
   Last version	Indicates the most recent version of the model.
   Deployment date	Displays the date when the deployment was initiated.
   Last status	Shows the last known status of the deployment process.

6. (Optional) Perform the following steps:
   1. Check for the latest version by clicking on the top-left corner of the console.
   2. Filter by parameters by clicking Status Filter from the list.
7. Review the deployment details, make sure everything is in order, and then click Deploy.

8. Observe the Operational status to see the deployment status.
   During the deployment of a trained Machine Learning model, you might see aby of the following status:

   PublishDeployCommand	The system is in the process of publishing the deploy command for the model.
   ModelDeployingAtEdge	The deployment process is currently underway.
   ModelDeployedAtEdge	The model has been successfully deployed.
   ModelDeploymentFailed	The deployment of the model encountered an error.

   If the status is ModelDeploymentFailed, you must take the following actions:

   • Review the configuration and ensure that all parameters are correctly set. To review the configuration file, go to Configuring-Machine-Learning-models.  
   • Verify that the model and its dependencies are compatible with the deployment environment.
   • If the issue persists, consider retraining the model or seeking assistance from technical support.
9. (Optional) To check the deployment status, click on the top-right corner of the console. 

To remove a trained Machine Learning model

You remove a trained Machine Learning model for several reasons to maintain the Machine Learning deployment's efficiency, relevance, and security. The following scenarios are when removing a trained Machine Learning model might be necessary:

• If a trained Machine Learning model no longer performs optimally or fails to meet the desired accuracy thresholds, you might remove the trained model to prevent inaccurate predictions or analyses.
• Older Machine Learning models might become obsolete as new data becomes available or business requirements evolve. You might replace it with updated versions to make sure relevance and accuracy.
• Removing redundant or unused Machine Learning models helps optimize computational resources, storage space, and memory, leading to cost savings and improved system performance.
• In situations where security vulnerabilities, such as susceptibility to adversarial attacks or data breaches, are identified in a trained machine learning model, remove the model to mitigate risks and protect sensitive information.
• Compliance with data protection regulations or industry standards might require the removal of Machine Learning models that violate privacy or confidentiality requirements, ensuring legal and ethical use of data.
• During system upgrades, migrations, or maintenance activities, removing outdated or unnecessary Machine Learning models streamlines the process and minimizes potential disruptions to operations.

The following video displays how to undeploy a trained Machine Learning model:

Use the following steps to remove a Machine Learning model. You use this step to stop the execution on specific nodes or devices. 

Before you begin

Make sure that you have deployed a trained Machine Learning model. 

1. On the BMC Helix Edge page, select Machine Learning.
2. Identify the specific trained Machine Learning model to remove from the list of trained models.
3. Click to expand the model console.
4. In the expanded trained Machine Learning console, under Operational status, look for models in a green deployment status.
5. Determine the trained model deployment status in green under the Operational status heading.  
6. Select Action > Undeploy. 
   BMC Helix Edge displays the list of nodes with various details related to the deployment.
7. Select a node that is attached to the trained model and click Undeploy.
8. (Optional) Click refresh on the console's top-left corner to check the deployment status.  

Where to go from here

Configuring-Event-generation-settings