This message was deleted.

Hey @Dean Shabi, you can achieve what you need in two ways: 1. use a custom Runnable where you load the latest

mlflow

model in the constructor of the runner. 2. use a custom Docker script where you modify the

ENTRYPOINT

to pull the latest model before running

serve

we have been receiving feedbacks from user that they would like to change the model without rebuilding the image. We are currently accessing several options and will mostly likely build something that would allow user to do this in the near future

That sounds perfect! Second option sounds good, and the upcoming feature would solve this need 🙂

you may look into some example here https://docs.bentoml.org/en/latest/guides/containerization.html