BentoML

Hi Amit, the following example make a custom runner that download the model from torchhub when start up. Maybe you can replace torchhub code with custom code downloading new model from S3?

<https://github.com/bentoml/BentoML/blob/main/examples/custom_runner/torch_hub_yolov5/service.py>

but, this will load it only on startup, no? which means i need to restart the service in order to reload the model

If you want to hot swap the model without restarting, I think one way to do is that create another method on the `Runnable` to reload the model, then create an extra http endpoint to trigger this reload method on the runnable