Model Deployment and Inferencing

Model Deployment

Once training completes, you can deploy the model in two ways:

• Deploy to Cortex: If you have registered a Cortex via the web console, simply click Deploy Artifacts to push the model. You must make sure the Cortex is in connected state and must be connected to internet.

• Download for External Use: If you prefer to deploy the model elsewhere, click to download the model artifacts directly to your local machine.

Model Inferencing

After you deploy the model on Cortex, you are ready to move back to AMAS for inferencing. Right below the interaction panel, you will find AI mode button. Select AI mode button to start model inferencing. You must make sure your task setup is ready before you enable AI mode.

Note: For ACT model, we currently enabled temporal_agg by default, so the robot execution might be slightly slow.

Human Intervention

Model takes control during inferencing, but you can take control back anytime by interacting with the green orb, just like how you normally control the robot in AMAS.

AI Mode Inferencing State Visual

When a robot enters AI Mode, to visualise its current state, its colour changes to yellow, and the mode indicator under the robot's name changes to Mode: Automated.