This message was deleted.

Here’s an example how to use “runner” in 1.2: https://github.com/bentoml/BentoSD2Upscaler/blob/main/service.py

In addition, 1.2 is fully backwards compatible with runners. Runner doc here: https://docs.bentoml.com/en/v1.1.11/

thanks! ill check this out

Thank you for the response. I'm interested in ControlNet with a runner implementation similar to the one described here: https://docs.bentoml.com/en/v1.1.11/frameworks/diffusers.html

@arnon_rodman, Hi! Here's an example project of controlnet deployment: https://github.com/bentoml/BentoControlNet

bento_model = bentoml.diffusers.get("sd2.1:latest")
sd21_runner = bento_model.to_runner(name="sd21-runner")

svc = bentoml.Service("stable-diffusion-21", runners=[sd21_runner])

with StableDiffusionXLControlNetPipeline

Hi All, I have a question about Runners. It was widely present in documentation 1.1, but totally absent in documentation 1.2. Does it mean that they will be deprecated in future o dramatically refactored?

@Vladislav Kievski I think it depends on the use case, but at the moment we're fully backwards compatible with runners. Were you using a custom runner? Or one that is out of the box?

We customized logic of bento 0.13 like InferenceAPI, Input/Output adapter, now we need to do something similar with bento 1. Probably it will require to customize some logic of Runners, so I am interested in future plans. If it will be changed in future, in this case we need to come up with better approach of integration without touching internals.