Auto-Sklearn estimators can not be cloned

[alexitkes]

Describe the bug

If I clone an auto-sklearn model, it crashes when fitting.

To Reproduce

Steps to reproduce the behavior:

from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split
from autosklearn.classification import AutoSklearnClassifier
from sklearn.base import clone

bc = load_breast_cancer()
(X, y) = (bc['data'], bc['target'])
(X_train, X_test, y_train, y_test) = train_test_split(X, y, random_state=0, stratify=y)

clf = AutoSklearnClassifier(time_left_for_this_task=45, per_run_time_limit=15, n_jobs=-1)
clf.fit(X_train, y_train)
print(clf.score(X_train, y_train))
print(clf.score(X_test, y_test))

m = clf.get_models_with_weights()[0][1]

m.fit(X_train, y_train)
print(clf.score(X_train, y_train))
print(clf.score(X_test, y_test))
# Output:
# 0.9906103286384976
# 0.9440559440559441

# Can a model be fitted again? OK.
m.fit(X_train, y_train)
print(clf.score(X_train, y_train))
print(clf.score(X_test, y_test))
# Output:
# 0.9906103286384976
# 0.9440559440559441

# Clone it
m = clone(m)

# And can a clone be fitted
m.fit(X_train, y_train)
print(clf.score(X_train, y_train))
print(clf.score(X_test, y_test))
# Crash!
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-7-6f7f5e57c192> in <module>
----> 1 m.fit(X_train, y_train)
      2 print(clf.score(X_train, y_train))
      3 print(clf.score(X_test, y_test))

~/venv-auto/lib/python3.6/site-packages/autosklearn/pipeline/base.py in fit(self, X, y, **fit_params)
     89             a classification algorithm first.
     90         """
---> 91         X, fit_params = self.fit_transformer(X, y, **fit_params)
     92         self.fit_estimator(X, y, **fit_params)
     93         return self

~/venv-auto/lib/python3.6/site-packages/autosklearn/pipeline/classification.py in fit_transformer(self, X, y, fit_params)
     96 
     97         X, fit_params = super().fit_transformer(
---> 98             X, y, fit_params=fit_params)
     99 
    100         return X, fit_params

~/venv-auto/lib/python3.6/site-packages/autosklearn/pipeline/base.py in fit_transformer(self, X, y, fit_params)
     99         fit_params = {key.replace(":", "__"): value for key, value in
    100                       fit_params.items()}
--> 101         Xt, fit_params = self._fit(X, y, **fit_params)
    102         if fit_params is None:
    103             fit_params = {}

~/venv-auto/lib/python3.6/site-packages/sklearn/pipeline.py in _fit(self, X, y, **fit_params)
    313                 message_clsname='Pipeline',
    314                 message=self._log_message(step_idx),
--> 315                 **fit_params_steps[name])
    316             # Replace the transformer of the step with the fitted
    317             # transformer. This is necessary when loading the transformer

~/venv-auto/lib/python3.6/site-packages/joblib/memory.py in __call__(self, *args, **kwargs)
    350 
    351     def __call__(self, *args, **kwargs):
--> 352         return self.func(*args, **kwargs)
    353 
    354     def call_and_shelve(self, *args, **kwargs):

~/venv-auto/lib/python3.6/site-packages/sklearn/pipeline.py in _fit_transform_one(transformer, X, y, weight, message_clsname, message, **fit_params)
    726     with _print_elapsed_time(message_clsname, message):
    727         if hasattr(transformer, 'fit_transform'):
--> 728             res = transformer.fit_transform(X, y, **fit_params)
    729         else:
    730             res = transformer.fit(X, y, **fit_params).transform(X)

~/venv-auto/lib/python3.6/site-packages/autosklearn/pipeline/components/data_preprocessing/data_preprocessing.py in fit_transform(self, X, y)
     88 
     89     def fit_transform(self, X, y=None):
---> 90         return self.fit(X, y).transform(X)
     91 
     92     @staticmethod

~/venv-auto/lib/python3.6/site-packages/autosklearn/pipeline/components/data_preprocessing/data_preprocessing.py in fit(self, X, y)
     77         self.column_transformer = sklearn.compose.ColumnTransformer(
     78             transformers=sklearn_transf_spec,
---> 79             sparse_threshold=float(self.sparse_),
     80             )
     81         self.column_transformer.fit(X)

TypeError: float() argument must be a string or a number, not 'NoneType'

Expected behavior

Work just like the model that was cloned

Environment and installation:

Ubuntu 18.04.4
Python 3.6.9
Auto-Sklearn 0.7.0

[mfeurer]

Fixed via #877. Could you please check if this works with the development branch, and reopen if this is not the case?