Question for python_deployer.deploy_python_closure - case of multiple outputs from func

[sean-sj-jung]

I am using Catboost (https://catboost.ai/) for classification and have a bit of complication with return values.

In case of xgboost, when you run xgboost.predict(xgb.DMatrix(x)), it returns a single value which gives the probability of a sample being in class 1 against class 0.

Catboost, however, returns both values (https://catboost.ai/docs/concepts/python-reference_catboostclassifier_predict_proba.html)
for example.
a, b = catboost_model.predict_proba(data)
where a = probability of being class 0, b = probability of being class 1,
and a+b should be equal to 1.

I defined my 'func' to be

def predict(x):
a, b = cat_clf.predict_proba(x)
return b

as I need the second value, the probability of being class 1, and put that in my model. The problem is, when I run the app, it somehow returns the value of 'a' instead of 'b'.
I saw your pytorch example (https://github.com/ucbrise/clipper/blob/develop/examples/frameworks/pytorch_on_docker_example.ipynb) and here you can see the identical behavior. cell number [7], the model returns two values but in cell number [21], the 'output' is the first value only.
How can I make it return the second value?