SAS Model Manager Overview
SAS Model Manager is a comprehensive platform designed to streamline and optimize the entire analytical modeling life cycle, from model creation and validation to deployment, management, governance, and monitoring. This flagship SAS product is tailored to support ModelOps activities, enabling organizations to efficiently manage and deploy machine learning models at scale.
Key Features and Functionality
Unify Model Assets
SAS Model Manager allows you to centralize and manage all your model assets in a secure, versioned model registry. This includes profiling, tagging, sorting, and categorizing models, as well as searching and discovering models across the organization. Users can access models and artifacts through the user interface or programmatically via REST API, SAS code, or Python code, ensuring seamless integration with automated MLOps processes. The platform supports models developed in SAS, Python, R, and other languages, and it provides auditing capabilities for major events like model creation, deletion, and deployment.
Validate Models
The platform ensures that models are thoroughly validated before deployment. It automatically generates scoring code for Python and R models and verifies that models run correctly within both SAS Model Manager and production environments. Users can compare SAS, Python, and R models side by side to determine the best fit for production, all from an easy-to-use, no-code interface.
Deploy Models in Minutes
SAS Model Manager facilitates rapid deployment of models across various environments, including cloud, on-premises, in-database, and within containers. Models can be deployed to destinations such as Snowpark, Azure Machine Learning, and Teradata without the need for recoding. The platform also supports deployment through SAS Cloud Analytic Services (CAS) and SAS Micro Analytic Service (MAS), which provide high-performance execution environments.
Monitor, Detect, Alert, and Repeat
The platform offers robust monitoring capabilities to detect data, concept, and model drift. It provides dynamic performance reporting from deployment to retirement and allows users to track, validate, and audit reports to select champion models. Customizable performance reports and Key Performance Indicators (KPIs) can be set up to alert stakeholders and cue model retraining, minimizing costly downtime.
Streamline MLOps Processes
SAS Model Manager integrates Continuous Integration and Continuous Delivery (CI/CD) processes to efficiently test and promote new models and decision flows. It uses templates to create repeatable workflows and automates the integration of multiple environments, tools, and applications through a flexible API ecosystem. This automation helps in keeping the right people informed and ensures that the workflow ecosystem is adaptable to different needs.
Additional Capabilities
- Workflow Management: The platform includes a workflow application that allows users to assign tasks, set up champion-challenger tournaments, and manage ModelOps activities such as importing models, setting and approving champion models, and publishing score code.
- KPIs and Performance Metrics: SAS Model Manager allows users to define, compute, and use standard and custom KPIs to track model performance over time. These KPIs can generate alerts when model performance exceeds threshold values, ensuring continuous model assessment and improvement.
- Model Governance and Transparency: The platform provides features to assess whether AI models are accurate, fair, and transparent using model cards, which is essential for model governance and compliance.
In summary, SAS Model Manager is a powerful tool that simplifies the management of machine learning models, ensuring they are developed, validated, deployed, and monitored efficiently. It integrates seamlessly with various environments and tools, making it an essential component for any organization engaged in advanced analytics and machine learning.