Azure Machine Learning - Public preview announcements for January 2023
发布日期：一月 18, 2023
New features now available in Public Preview include the ability to build an end-to-end model training pipeline without needing to write any code, track whether a new version of your compute instance is available, recover a workspace to ensure little disruption in your work operations, create schedules to automatically run jobs on a regular basis, and gain insight into potential root causes of pipeline failure.
- Train models using AutoML in Designer: You can now drag-and-drop your data, hook it up to an AutoML task and deploy the resulting best model. This is the perfect solution for no-code/low-code users who want to build enterprise-quality machine learning models.
- Audit and observe compute instance OS version: You can now track your compute instance’s security patch compliance and receive notifications when updates are needed.
- Recover deleted workspace data by enabling soft delete: You can now work with the assurance of stronger data workspace protection and minimal disruption when recovering a workspace that had been soft deleted.
- Create and manage your schedules in AzureML Studio: You can now easily create a schedule based on a pipeline job, without the limitations of only using Python SDK.
- Compare different pipelines to debug failure: You can now save debugging time with new insight into why a specific pipeline may have failed.
- View profiling to debug pipeline performance issues: You can now identify which nodes are problematic and which are unproblematic in order to save debugging time.