Closed
Description
After upgrading to 0.12.0, I get the warning before training starts:
UserWarning: You set iterator_valid__shuffle=True; this is most likely not what you want because the values returned by predict and predict_proba will be shuffled.
However, I definitely create the model with iterator_valid__shuffle = False
. And, just to confirm, after training I do
print(estimator.get_params()["iterator_valid__shuffle"]) # False
I'm hoping the warning itself is wrong, otherwise I don't see how I can fix this within my own code.