Why lime tabular method generates type errors with data frames?

Question

I'm traying to use LIME to explain the results of a gradient boost model.

This is the data I have:

And the code I'm using:

explainer = lime.lime_tabular.LimeTabularExplainer(training_data = sample, 
                                                   mode = 'regression',
                                                   training_labels = y_train,
                                                   feature_names = ["x1","x2"])

And the error I´m getting:

TypeError: unhashable type: 'slice'

Answer 1

You probably should convert sample to a numpy array.

From the source code:

training_data: numpy 2d array