Deployments

A deployment is essentially an active model; if a model is deployed, inferences can be tested against it. After a model has been deployed, an entry for that deployment will be available on the Deployments page. AAW currently supports three types of deployments: On Demand, Batch, and Continuous. On Demand deployments can be tested manually and as necessary with user input. Batch deployments inference on a batch of data all at once. Continuous deployments utilize table monitors to convert streaming incoming data into continuous inferencing output.

Important

Models are stopped by default after deploying for the first time.

• Type into Filter to filter down the deployments
• Click to display any archived deployments.
• Click to refresh the table
• Click to export the table's values as JSON or CSV
• Click an existing deployment to display additional actions. Actions vary depending on the type of model:
  • Continuous / Batch deployments:
    • Click View Selection to open the Deployment Details page
    • Click View Inferences to view the inference results for the selected deployment
    • Click Export Entity to export the deployment as a JSON object
    • Click Start to start the deployment
    • Click Stop to stop the deployment
    • Click Archive to archive the deployment; it will be hidden from the list of deployments
    • Click Description / Config to review summary information for the deployment
  • On-demand deployments:
    • Click View Selection to open the Deployment Details page
    • Click Test Inference to test the model using manual input
    • Click Export Entity to export the deployment as a JSON object
    • Click Start to start the deployment
    • Click Stop to stop the deployment
    • Click Archive to archive the deployment; it will be hidden from the list of deployments
    • Click Description / Config to review summary information for the deployment

Details

The Deployment Details page provides a detailed look at a given deployment, including deployed models' replicas' log access, table monitor information, and state. Available actions depend on the type of model being detailed.

• Continuous / Batch deployments:
  • Click Back to return to the Deployments page
  • Click View Inference to view the inference results for the selected deployment
  • Click Export Entity to export the deployment as a JSON object
  • Click Start to start the deployment
  • Click Stop to stop the deployment
  • Click Archive to archive the deployment; it will be hidden from the list of deployments
  • Click View Log to view a given replica's log
• On-demand deployments:
  • Click Back to return to the Deployments page
  • Click Test Inference to test the model using manual input
  • Click Export Entity to export the deployment as a JSON object
  • Click Start to start the deployment
  • Click Stop to stop the deployment
  • Click Archive to archive the deployment; it will be hidden from the list of deployments
  • Click View Log to view a given replica's log