In the “Autoscaling TensorFlow model deployments with TF Serving and Kubernetes”, the command
used to get a prediction (also to do the stress test) fails with this:
Servable not found for request: Latest(image_classifier)
Doing some investigation it looks like the environment variables MODEL_NAME and MODEL_PATH are empty even if I manually set them as “image_classifier” and to my cloud bucket in both the deployment and config YAML
When I do the stress test, this error doesn’t “stress” the system at all. Any hints on what’s happening here?