BentoML

It works on cpu’s on my local machine though.

Added debug logging to the service as well, but there are no debug logs still.

```@svc.on_deployment
def startup_service():
    print(f"Initializing {SERVICE_NAME} with {MODEL_NAME}:{MODEL_VERSION}...")
    stream_handler = logging.StreamHandler()
    formatter = logging.Formatter(
        "%(asctime)s - %(name)s - %(levelname)s - %(message)s"
    )
    stream_handler.setFormatter(formatter)
    bentoml_logger = logging.getLogger("bentoml")
    bentoml_logger.addHandler(stream_handler)
    bentoml_logger.setLevel(logging.DEBUG)```

Hey <@U05HV760F9C>, could you share your runner and service code?

Figured out why the logs in the `__init__` if the model weren’t running. I think it’s because bento pickles the initialized model when we call `save_model` so it doesn’t have to re-init each time it schedules it on a runner. Does that seem right?