Issue with Bokeh and interactive classifier

[ebbestubbe]

I was following the calmcode.io draw tutorial here
https://calmcode.io/human-learn/draw.html

Working with Bokeh was a bit clunky(did not work in vscode or firefox, and it was quite slow in Chrome), but I did end up getting the json of the hulearn.experimental.interactive.InteractiveCharts object.
attached here: penguin_draw.json
In the fitting/drawing part of the demonstration, preds = model.fit(X, y).predict_proba(X) I get this error:

Traceback (most recent call last):
  File "C:\Users\stubb\AppData\Local\Temp\ipykernel_22968\2726856544.py", line 9, in <module>
    preds = model.fit(X, y).predict_proba(X)
  File "C:\repos\titanic\venv\lib\site-packages\hulearn\classification\interactiveclassifier.py", line 148, in predict_proba
    hits = [
  File "C:\repos\titanic\venv\lib\site-packages\hulearn\classification\interactiveclassifier.py", line 149, in <listcomp>
    self._count_hits(self.poly_data, x[1].to_dict()) for x in X.iterrows()
  File "C:\repos\titanic\venv\lib\site-packages\hulearn\classification\interactiveclassifier.py", line 109, in _count_hits
    point = Point(data_in[c["x_lab"]], data_in[c["y_lab"]])
KeyError: 'bill_length_mm'

This is my environment. I couldn't find a requirement file for the examples provided, so I had to install the packages by the latest versions(as instructed in the examples)
reqs.txt

(Btw thanks for a quick fix on the last issue, I was too slow to submit my own PR over the holidays :) )

[koaning]

It might be that there's an issue related to the Bokeh upgrade a year ago. I think there were some breaking changes from v1 -> v2.