feature_importance causes a BSOD on Windows 10

Describe the bug

Running permutation_importance on a medium-sized data set results in a BSOD on Windows 10. The dataset is 470605 x 332, code is running in a Jupyter notebook, Python version 3.7.6, scikit version 0.22.1. The BSOD is a KERNEL_SECURITY_CHECK_FAILURE, with ERROR_CODE: (NTSTATUS) 0xc0000409 - The system detected an overrun of a stack-based buffer in this application. This overrun could potentially allow a malicious user to gain control of this application. The machine has a Ryzen 5 3600 with 16GB of RAM.

Steps/Code to Reproduce

from sklearn.ensemble import RandomForestClassifier
from sklearn.inspection import permutation_importance
rf = RandomForestClassifier(n_estimators = 250,
                           n_jobs = -1,
                           oob_score = True,
                           bootstrap = True,
                           random_state = 42)
rf.fit(X_train, y_train)
permImp = permutation_importance(rf,
                                 X_val,
                                 y_val,
                                 scoring='f1',
                                 n_repeats=5,
                                 n_jobs=-1,
                                 random_state=42)

Expected Results

No BSOD, permutation importance computed.

Actual Results

BSOD after ~1-2 minutes

Versions

sklearn.show_versions()

System: python: 3.7.6 (default, Jan 8 2020, 20:23:39) [MSC v.1916 64 bit (AMD64)] executable: C:\Users\lucag\anaconda3\python.exe machine: Windows-10-10.0.18362-SP0

Python dependencies: pip: 20.0.2 setuptools: 45.2.0.post20200210 sklearn: 0.22.1 numpy: 1.18.1 scipy: 1.4.1 Cython: 0.29.15 pandas: 1.0.1 matplotlib: 3.1.3 joblib: 0.14.1

Built with OpenMP: True

[ogrisel]

Actually, looking at the code, using mmaping should be fine (and useful) because joblib.Parallel is called in the non permuted data and the permutation is applied on a copy of the data in the worker.

Maybe there is something wrong and low level only happening on windows. I need to investigate with a windows VM.

[Devilmoon]

Unfortunately I already removed those folders and cannot remember the exact file names, however I can tell you that they all had short file names, and all of them had the same size, which was around ~2MB