Questo contenuto non è disponibile nella lingua selezionata.
Chapter 2. New features and enhancements
This section describes new features and enhancements in Red Hat OpenShift AI.
2.1. New Features
- Tabular data drift detection with KServe
As part of ongoing advancements in the Responsible AI space, backend metrics support for drift monitoring is now available in the TrustyAI project. These capabilities align with the TrustyAI framework to ensure secure and reliable AI operations.
Drift monitoring is crucial for maintaining AI model performance and fairness over time. By implementing backend metrics first, this release lays the groundwork for comprehensive drift detection and monitoring, enhancing the trustworthiness and compliance of AI deployments. Future updates will extend these capabilities with a user interface for streamlined interaction.
- Bias and fairness monitoring
- Using TrustyAI, data scientists can now evaluate deployed models for bias and fairness. The update includes both backend functionality and dashboard visualizations.
- Support heterogeneous clusters in distributed workloads
Data scientists can now select specific queues based on their workload requirements, improving efficiency in resource-constrained environments. Administrators can configure these workload queues per cluster and data science project.
Additionally, the dashboard now supports visibility across all queues in heterogeneous clusters. You can view both the assigned queue and resource details for each workload, providing clear insights into workload distribution.
2.2. Enhancements
- Data science pipelines updates
The data science pipelines components are updated to Kubeflow Pipelines 2.2.0. For more information, see the version 2.2.0 changelog.
Additionally, the internal Argo workflows engine was updated to version 3.4.17 and includes security fixes.
