Hello, is it possible to use bentoml service as training process API, where API endpoint receives data as. json file and store trained model in bento Model store?

It is possible and we usually recommend using the

@task

abstraction for this. https://docs.bentoml.com/en/latest/guides/services.html#tasks You may also want to limit the concurrency to 1 in case of OOM issues. https://docs.bentoml.com/en/latest/bentocloud/how-tos/autoscaling.html#concurrency

@Chaoyu Thanks for the quick response! So based on the Task document, I'm guessing the barebone API code structure would look like following?

import bentoml

@bentoml.service(
    traffic={
        "concurrency": 1,
    }
)
class MachineLearningService:

    @bentoml.task
    def get_train_data_train_model(self, data: dict) -> dict:
        # Process training
        return data
    
    @bentoml.api
    def predict(self, data: dict) -> dict:
        # Process predict
        return data

yes looks right!