This message was deleted.

Hello again, when I trying to find a solution for that I found custom_runner.init_local() function that makes runner work on api request. However, when I try to call custom_runner.init_client() it does not work because of name error. Is there anyone who has a solution for this or approriate approach to create runners at runtime. Thank you

hello @FIRAT TAMUR have you get a chance to look at this doc: https://docs.bentoml.org/en/latest/guides/graph.html#inference-graph

Hello, First of all, thank you for the answer and this great framework. Actually, I read this page and looked at it again. But it does not fit my problem. In this example or examples in the doc, runners are defined before the service is created but I want to create a new runner depending on the request in the defined service. For example, I have 5 models in my ./models folder, I made a service with 2 of the models and started the server. I want to load another model from my other 3 models if any related request comes. Thanks again for support 🙂