Closed
Description
Description
In version 0.18, loading pickles of fitted RandomizedSearchCV objects results in a TypeError exception (from pickle also created with version 0.18).
The error seems related to the use of masked arrays in the RandomizedSearchCV.cv_results_ attribute - clearing this before pickling (i.e. setting to to None) allows pickling/unpickling to work.
Steps/Code to Reproduce
import pickle
from sklearn.model_selection import RandomizedSearchCV
from sklearn.ensemble import RandomForestClassifier
X = [[1, 0], [1, 0], [1, 0], [0, 1], [0, 1], [0, 1]]
y = [1, 1, 1, 0, 0, 0]
model = RandomizedSearchCV(RandomForestClassifier(),
{'n_estimators': [5, 10, 20]},
n_iter=3)
model.fit(X, y)
with open('model.pkl', 'wb') as fh:
pickle.dump(model, fh)
with open('model.pkl', 'rb') as fh:
model = pickle.load(fh)
print(model.predict(X))
Expected Results
[1, 1, 1, 0, 0, 0]
Actual Results
Traceback (most recent call last):
File "./t.py", line 19, in <module>
model = pickle.load(fh)
File "/Users/dsc/miniconda3/envs/p3/lib/python3.5/site-packages/numpy/ma/core.py", line 5863, in __setstate__
super(MaskedArray, self).__setstate__((shp, typ, isf, raw))
TypeError: object pickle not returning list
Versions
Python 3.5.1 |Continuum Analytics, Inc.| (default, Dec 7 2015, 11:24:55)
[GCC 4.2.1 (Apple Inc. build 5577)]
NumPy 1.11.1
SciPy 0.18.1
Scikit-Learn 0.18