Release Notes

v3.2.0

• Endpoints update:
  • New Endpoint type: Kafka endpoints with real time streaming.
  • New Endpoint type: Batch prediction endpoints.
  • Canary release overhaul.
  • Feedback Loops for exporting endpoint data.
  • Create live dynamic endpoint charts.
• New AI Library, Python SDK and CLI updates to support the new features.
• Various bug fixes and optimizations.

v3.1.0

• Human Validation: Tasks in flows can now optionally have approval required. These tasks will not run until approval is received.
• Experiment Visualization update: Steps have been added to experiment visualizations, allowing you to chart multiple matrices and bar graphs over time in a single graph.
• Various bug fixes and optimizations.

v3.0.2

• Flow YAML update: You can now add scheduling information in Flow YAMLs.
• R Shiny update: You can now use custom docker images with R Shiny dashboards.
• Email notification update: Support for emailing multiple people for experiment status notifications.
• Various bug fixes and optimizations.

v3.0.1

• Stability improvements.
• Various bug fixes and optimizations.

v3.0.0

• cnvrg CORE has been released: Deploy it for yourself to get started.
• New workspace type: Visual Studio Code. You can now select Visual Studio Code as a workspace type.
• Open MPI support: You can now run distributed jobs over multiple nodes using Open MPI.
• Compute Resources overhaul:
  • The system to add a new Compute Resource (Kubernetes cluster or on-premise machine) has been refactored and streamlined.
  • Added Compute Heath Checks: cnvrg will automatically check your compute resource is online and update you to its status.
  • Added System Dashboard: Use the new built-in dashboard to monitor utilization with GPU, CPU, Memory, Temperature and more metrics.
  • Added other self monitoring dashboards: Use Grafana and Kibana to administrate connect Kubernetes clusters.
• New deployment for cnvrg and cnvrg CORE: the process to deploy cnvrg has been made easier through the use of Helm 3.0.
• Onboarding update: The onboarding for new users has been rebuilt for the personalization of your cnvrg experience.
• Flows update: Flows now run as a DAG.
• Python SDK updated: Support for new features and stability improvements (v0.1.9).
• cnvrg CLI updated: Support for new features and stability improvements (v1.6.32).
• Various bug fixes and optimizations.

Pre-3.0.0

• The container management page has been relocated. It is now a tab on the organizational level.
• NFS support: You can now cache dataset commits to an external NFS disk and make commits available immediately with no cloning.
• Job History update: Go to Compute > Jobs to see the new job history viewer.
• Overhaul of Docker image/container management: You can now pull/build and use any custom docker image from any registry.