Now that we've built a model, let's deploy a hosted SageMaker inference endpoint and use it to get some predictions.
Download the template file for an endpoint. Make the following changes:
kubectl describe trainingjob <JOBNAME>and looking for the
kubectl apply -f tf-endpoint.yaml
You can check on the endpoint status by running
kubectl get hostingdeployments and
kubectl describe hostingdeployments <ENDPOINT>.
Find the IAM role whose name starts with
eksctl-eks-kubeflow-nodegroup. Attach the policy
AmazonSageMakerFullAccess to let this role invoke our endpoint.
Download the example notebook. In your notebook, click
Upload and select this notebook. Click
Upload again to complete the upload.
Now click the imported notebook to open it. Read through and execute each cell in the notebook, making sure to edit the variables in the first code cell. You can find the
endpoint by describing your inference endpoint.
After you finish executing this notebook, you'll see metrics like precision and recall for our model.