Is it possible to pass init parameters to Predictor class through seldon-core-microservice

[zyxue]

my custom model is like:

class MyPredictor:
    def __init__(self, model_path):
        self._model = load(model_path)

    def predict(self, features, names, meta=None):
        transformed = transform(features)
        prediction = calc_prediction(transformed)
        return prediction

I wonder if there is a way to pass value to model_path through seldon-core-microservice?

Went throught the source code, https://github.com/SeldonIO/seldon-core/blob/master/python/seldon_core/microservice.py#L212-L296, there doesn't seem to be one. Am I missing some thing or is there another way to do something like this?

[ukclivecox]

Are you looking for https://docs.seldon.io/projects/seldon-core/en/latest/python/python_wrapping_s2i.html#model-class-arguments

[zyxue]

yes, this is what I'm looking for

• https://github.com/SeldonIO/seldon-core/blob/master/python/seldon_core/microservice.py#L222-L224,
• https://github.com/SeldonIO/seldon-core/blob/master/python/seldon_core/microservice.py#L68-L114

I didn't fully get how --parameters is used, now it makes sense. Thank you!

[nishantgandhi99]

In case someone come looking for example:

$ seldon-core-microservice InferenceModel --service-type MODEL --debug --parameters '[{"name":"model_path", "value":"value", "type": "STRING"}]'