Skip to main content

Model Deployment

Deploy the checkpoint selected during real-world validation, then verify it on the host that will execute the task.

Deployment sequence​

  1. Identify the accepted checkpoint in Model Management. Record its dataset, policy, and training step.
  2. Check the policy and target-device combination in Policy Selection.
  3. Where the selected device requires conversion, follow Inference & Model Conversion and evaluate the converted artifact.
  4. If the target is a different host, follow Cross-Host Transfer & Deployment Planning to share and receive the selected model.
  5. On the target host, confirm the physical arm and cameras, select the received model, and repeat the real-world validation.

For a single-host setup, use the selected checkpoint on that host; cross-host transfer is only needed when assets move between hosts.

TaskGuide
Configure the physical setupHardware preparation
Review, merge, or share datasetsDataset management
Select, import, or share checkpointsModel management
Plan hosts and transfer assetsCross-host transfer

Retain the model identification and verification results with the deployment record. If camera placement, objects, or task conditions change, repeat validation and use the diagnosis guide before choosing the next corrective step.

Return to the Sim to Real overview, or consult the FAQ.