AI Fabric is a service that allows you to deploy and manage Machine Learning models and consume them within RPA workflows in Studio. The ML models can be built in a Python IDE or using an AutoML platform such as H20 Driverless AI. This chapter treats the subject of model deployment and management which is done at the tenant level in Orchestrator.
AI Fabric is only available for Cloud Platform tenants and is licensed as a separate service.
- To gain access to AI Fabric, request and then allocate AI Robot licenses to your tenant within your Cloud Platform account.
The ML-pertaining options are displayed in Orchestrator if your tenant has the necessary licenses and you have the corresponding permissions.
In order to perform various actions regarding ML entities, you need certain permissions:
- Display ML logs and view log data - View on ML Logs
- Display ML packages and view package details (available versions, parameters, change log) - View on ML Packages
- Displays ML skills and allows you to view details on the corresponding ML package (available versions, parameters) - View on ML Skills
- Upload a new ML package. - Create on ML Packages.
- Deploy a new ML skill - Create on ML Skills.
- Upload a new ML package version and view release notes for each version - Edit on ML Packages
- Update to a new ML skill package version or rollback to an older one. - Edit on ML Skills
- Remove undeployed package versions. - Delete on ML Packages
- Remove ML skills. - Delete on ML Skills
This section treats the subject of user personas which handle AI Fabric in your company, and the recommended roles to be defined in Orchestrator for each persona.
In charge of building and uploading the ML models to AI Fabric. Data Scientists build and then upload ML packages to Orchestrator. They can perform this operation from the ML Packages page in Orchestrator.
- View, Edit, Create, Delete on ML Packages
- View on ML Skills
- View on ML Logs
In charge of deploying models already uploaded by Data Scientists (ML packages), or provided by UiPath (OS packages), into ML skills within their Orchestrator tenant. Process Controllers can perform this operation from the ML Skills page in Orchestrator.
- View, Edit, Create, Delete on ML Skills
- View on ML Packages
- View on ML Logs
In charge of developing and testing automation workflows; usually does not have access to Orchestrator. RPA Developers consume the ML skills available on their Robot. These are retrieved from the Orchestrator tenant the connected Robot has been provisioned in.
- View on ML Skills.
- View on ML Logs.
RPA Developers consume deployed ML skills within customized workflows in Studio using the ML Skill activity from the UiPath.MLServices.Activities package. This activity package is only available for Studio v2019.10+ and can only be used by Robots v2019.10+. Read more about it here.
Updated about a month ago