BentoML

Hi, is Pytorch no longer natively supported in the latest version of BentoML? I don't see any examples available
<https://docs.bentoml.com/en/latest/examples/overview.html>

As pytorch is generally a very flexible framework, we have deprecated the support of `bentoml.pytorch`  to give user more flexibility. To save your pytorch model with bentoml, we would point you to using the `bentoml.models.create` API
<https://docs.bentoml.com/en/latest/build-with-bentoml/model-loading-and-management.html>

which allow you to pack any arbitrary files into a "Model"

Hi <@U03UY1JB269>, I'm trying to dynamically add multiple models into the same BentoML service.

Basically in `service.py`, I'm trying to load a list of BentoML models into a single variable. But when I do `bentoml build` and then `bentoml export` the .zip file doesn't contain the models.

It only works when each BentoML model is a variable.

I've replicated the unsuccessful code using the XGBoost example: <https://github.com/TiewKH/BentoXGBoost/tree/main>

May i know if this is expected?

I think this is expected behaviour. packaging model into the bento works by scanning class variable of type `BentoModel`. <@U054RJZBWFJ> could you confirm?

This is the user's code
```class CancerTiewKHClassifier:
    # Retrieve the latest version of the model from the BentoML model store

    # This fails
    bento_model = [BentoModel(f"{model_name}:latest")]

    # This works
    # bento_model = BentoModel(f"{model_name}:latest")```