Skip to main content
You train a policy in the dashboard. It runs on a GPU for you. No code needed. You need a finished dataset first. This is step 3 of Train a policy.
1

Click Train

Go to Data → Datasets, open your dataset, and click Train.
A dataset page with the Train button
2

Fill in the panel

The Train a policy panel
Click Train. You can close the panel. Training keeps going.
3

Watch the progress

Go to AI → Training and open your run. It goes through these stages:Queued → Initializing → Preparing Dataset → Downloading Data → Validating Data → Training → Saving Checkpoints → CompletedThe page also shows training metrics and saved checkpoints. If training fails, the reason is shown at the top.
4

Copy the Training ID

When the run says Completed, click the copy button next to the Training ID. It looks like 123e4567-e89b-42d3-a456-426614174000.
A completed training run with the Training ID and its copy button
Next step: run the policy with this Training ID.

Running a policy

Next: let the policy drive the arm.

Datasets

Back: make a dataset.