Managing ML Skills

Creating ML Skills

1. In the ML Skills page, click Create New. The Create New ML Skill page is displayed.
2. Enter a name for the new skill. The name should contain only characters, numbers, and underscores and should not start with number.
3. Select a package from the drop-down.
4. Select package major and minor versions.
5. Optional: Add a description for the new skill.
6. Choose whether or not to enable a GPU on the environment running this skill
7. Choose whether or not to enable auto-update. If this is set to true, the model is automatically updated with the latest retrained version of the model (same ML package version, but new minor version).
8. Choose when to undeploy the Skill after a period of inactivity from the drop-down. A Skill is considered as inactive if it is not used at all and there are no predictions or changes on the deployment itself.
   
   

   

   
   
9. Choose whether or not to make the ML Skill public.
10. Optional: Toggle the Enable Infra Settings if you want to customize the resources and replicas for this specifc environment. If this is enabled, the following fields can be customized:
    • Number of replicas.
    • The request and limit sizes of RAM (GB).
    • The request and limit sizes of CPU.
    Note: Only change these values if you are sure they are the correct ones.
    
    

    

    
    
11. Click Create. The Create New ML Skill page is closed and the ML Skills page is displayed, containing the new skill with Deploying status.
    
    

    

    
    
    The model is wrapped in UiPath's serving framework and deployed within a namespace on UiPath AI Center™'s Kubernetes cluster that is only accessible by your tenant.

    Note: Model deployment usually takes betwen 10-15 minutes.

    If the deployment is successful, the status of the skill changes from Deploying to Available.

Stopping ML Skills

To stop an ML Skill, use the following procedure:

1. In the ML Skills page, click on the skill you want to stop.
2. Click on the Stop button from the top.

Redeploying stopped ML Skills

To redeploy an ML Package, use the following procedure:

1. In the ML Skills page, click on the stopped Skill on the grid.
2. Click on the Resume button from the top.
   
   

   

   
   

Managing package versions

The ML Skill Details page enables you to manage the package version used within the deployed skill. You can update the skill to use a specific package version, or you can perform a rollback to the previously used version.

Updating package version

1. In the ML Skill Details page, click ⇅ next to the package version you want to use. The button is disabled for the version that is currently in use, marked as Current. The Update skill window is displayed.
2. In the Update skill window, click Confirm to update the skill to use the selected package version.
   
   

   

   
   

Rolling back to previous version

1. In the ML Skill Details page, click Rollback. The Update Skill window is displayed.
2. In the Update Skill window, click OK to update the skill to use the previous used package version.

Modifying Current Deployment

In addition to updating a skill to a different version, a deployed skill can be kept at the same version but modified to use or not use a GPU. To do that, select Modify current deployment in the ML Skill Details page > Update skill window, as below:

Make ML Skill public/private

You have the possibility to make an ML Skill public. Doing so, the ML Skill will be accessible via an endpoint from outside of UiPath environment. This means that you can call it without the need to go through a robot connected to the specific tenant.

To do that, select Modify current deployment in the ML Skill Details page > Update skill window.

Note this will redeploy the ML Skill, once the ML Skill is available you will see the corresponding url in the ML Skill Details and a button will help you copying the URL. Depending of base ML Package an API Key will also be exposed to be used with this Skill in ML Skill activity.

Deleting ML Skills

1. In the ML Skills page, click the Cancel or Remove button, depending on the status, next to a deployed or failed skill. A confirmation window is displayed.
   Depending on the ML Skill status, use one of the following buttons to delete it:

   • Cancel: use this button to delete a failed ML Skill.
   • Remove: use this button to delete a successfully deployed ML Skill.
2. Click OK to remove or cancel the skill. The selected skill is undeployed and it disappears from the ML Skills page.
   Note: Deleting a skill could affect its consumers. However, a skill with the same name can be re-created at any point restoring dependents of this service.