New features now in public preview include the functionality to promote pipelines and models across AzureML Workspaces and shorten training times, and lower set-up costs with the newest available, and compatible versions of training software for PyTorch.
- Use AzureML registries to share ML pipelines and models across teams and workspaces: This feature enables you to access cross-workspace collaboration across MLOps. Data science teams within an organization can now work together to create registries, publish pipelines and models, and promote models across development, test, and production environments.
- Optimize training with the Azure Container for PyTorch: This feature is the latest PyTorch environment, and it is optimized for large, deep-learning workloads. This lightweight, standalone environment offers you accelerated training at reduced costs and in less time.