I'm upgrading from v1.1 to v1.3 and I can't figure out how to achieve the following. I have customized the

bentoml.Service

class - so my code at the moment looks like

svc = MyCustomService(...)

@svc.api(input=..., output..., doc=...)
async def classify(...)

    return ...

I wonder, now with the class based services, I guess that I can still wrap the methods with the save decorator - but I'm wondering how would I connect it to the class itself? For exmaple, from your XGBoost docs:

@bentoml.service(
    resources={"cpu": "2"},
    traffic={"timeout": 10},
)
class CancerClassifier:
    # Retrieve the latest version of the model from the BentoML Model Store
    bento_model = bentoml.models.get("cancer:latest")

    def __init__(self):
        self.model = bentoml.xgboost.load_model(self.bento_model)

        # Check resource availability
        if os.getenv("CUDA_VISIBLE_DEVICES") not in (None, "", "-1"):
            self.model.set_param({"predictor": "gpu_predictor", "gpu_id": 0})  # type: ignore (incomplete XGBoost types)
        else:
            nthreads = os.getenv("OMP_NUM_THREADS")
            if nthreads:
                nthreads = max(int(nthreads), 1)
            else:
                nthreads = 1
            self.model.set_param(
                {"predictor": "cpu_predictor", "nthread": nthreads}
            )

    @bentoml.api
    def predict(self, data: np.ndarray) -> np.ndarray:
        return self.model.predict(xgb.DMatrix(data))

How would I replicate this with a custom service?

Hello @Elior Cohen, i am not sure if I understood your question, are you looking to create a base class?

mmm

oh, you can do

@bentoml.service
class CustomService:
    ...


class ChildService(CustomService):
    ...

@Jian Shen Yap I did exactly that - now using my new childservice - how do I apply it this example

@bentoml.service(
    resources={"cpu": "2"},
    traffic={"timeout": 10},
)
class CancerClassifier:
    # Retrieve the latest version of the model from the BentoML Model Store
    bento_model = bentoml.models.get("cancer:latest")

    def __init__(self):
        self.model = bentoml.xgboost.load_model(self.bento_model)

        # Check resource availability
        if os.getenv("CUDA_VISIBLE_DEVICES") not in (None, "", "-1"):
            self.model.set_param({"predictor": "gpu_predictor", "gpu_id": 0})  # type: ignore (incomplete XGBoost types)
        else:
            nthreads = os.getenv("OMP_NUM_THREADS")
            if nthreads:
                nthreads = max(int(nthreads), 1)
            else:
                nthreads = 1
            self.model.set_param(
                {"predictor": "cpu_predictor", "nthread": nthreads}
            )

    @bentoml.api
    def predict(self, data: np.ndarray) -> np.ndarray:
        return self.model.predict(xgb.DMatrix(data))

There are two usages here of the service the

@bentoml.service

and the

@bentoml.api

- how do I apply the child service I created to the class and method decorators?

Hello, sorry i let this slip past my backlog. Maybe i did not understood your use case completely, would doing

class ChildClassifier(CancerClassifier):
    ...

works for you?

Okay - and now how 1. How do pass

resources

and

traffic

as in the XGBoost example (the class level decorator) 2. How do I decorate the

predict

function which now has

@bentoml.api

- what should I put instead of it? Because internally

@bentoml.api

creates a different

Service

and we want to use the new one (

ChildClassifier

)