This accelerator is not expected to run out-of-box without some intervention

[dbbnicole]

This accelerator uses model serving.
The user must manually turn on classic model serving before the inference code and be run successfully.