Segfault in HistGradientBoostingClassifier

Describe the bug

I trigger a segfault in HistGradientBoostingClassifier. ~~I could trigger during cross-validation with n_jobs=-1 and n_jobs=1.~~Actually, I am not able to trigger anymore in n_jobs=1 but it was the case before (on a case without a random_state set.

I am using both missing values and categorical features management at the same time. I don’t know if it could be one of the issue.

Steps/Code to Reproduce

# %%
import pandas as pd

target_name = "RainTomorrow"
data = pd.read_csv("./weather.csv", parse_dates=["Date"])
data = data.dropna(axis="index", subset=[target_name])
X, y = data.drop(columns=["Date", target_name]), data[target_name]

# %%
X.info()

# %%
from sklearn.preprocessing import OrdinalEncoder
from sklearn.compose import make_column_transformer, make_column_selector

categorical_columns = make_column_selector(dtype_include=object)(X)
preprocessing = make_column_transformer(
    (
        OrdinalEncoder(handle_unknown="use_encoded_value", unknown_value=-1),
        categorical_columns,
    ),
    remainder="passthrough",
)

# %%
from sklearn.pipeline import make_pipeline
from sklearn.ensemble import HistGradientBoostingClassifier

model = make_pipeline(
    preprocessing,
    HistGradientBoostingClassifier(
        categorical_features=range(len(categorical_columns)),
        random_state=0,
    ),
)

# %%
from sklearn.model_selection import cross_validate

cross_validate(model, X, y, n_jobs=-1)

I am also attaching the dataset that I used to trigger the problem.

weather.csv

I tried to reproduce with a random set with both categorical and missing values but it did segfault.

Expected Results

At least it should not segfault.

Actual Results

---------------------------------------------------------------------------
TerminatedWorkerError                     Traceback (most recent call last)
~/Documents/scratch/bug_hist_gradient_boosting.py in <module>
      40 from sklearn.model_selection import cross_validate
      41 
----> 42 cross_validate(model, X, y, n_jobs=-1)

~/Documents/packages/scikit-learn/sklearn/model_selection/_validation.py in cross_validate(estimator, X, y, groups, scoring, cv, n_jobs, verbose, fit_params, pre_dispatch, return_train_score, return_estimator, error_score)
    265     # independent, and that it is pickle-able.
    266     parallel = Parallel(n_jobs=n_jobs, verbose=verbose, pre_dispatch=pre_dispatch)
--> 267     results = parallel(
    268         delayed(_fit_and_score)(
    269             clone(estimator),

~/Documents/packages/joblib/joblib/parallel.py in __call__(self, iterable)
   1052 
   1053             with self._backend.retrieval_context():
-> 1054                 self.retrieve()
   1055             # Make sure that we get a last message telling us we are done
   1056             elapsed_time = time.time() - self._start_time

~/Documents/packages/joblib/joblib/parallel.py in retrieve(self)
    931             try:
    932                 if getattr(self._backend, 'supports_timeout', False):
--> 933                     self._output.extend(job.get(timeout=self.timeout))
    934                 else:
    935                     self._output.extend(job.get())

~/Documents/packages/joblib/joblib/_parallel_backends.py in wrap_future_result(future, timeout)
    540         AsyncResults.get from multiprocessing."""
    541         try:
--> 542             return future.result(timeout=timeout)
    543         except CfTimeoutError as e:
    544             raise TimeoutError from e

~/mambaforge/envs/dev/lib/python3.8/concurrent/futures/_base.py in result(self, timeout)
    442                     raise CancelledError()
    443                 elif self._state == FINISHED:
--> 444                     return self.__get_result()
    445                 else:
    446                     raise TimeoutError()

~/mambaforge/envs/dev/lib/python3.8/concurrent/futures/_base.py in __get_result(self)
    387         if self._exception:
    388             try:
--> 389                 raise self._exception
    390             finally:
    391                 # Break a reference cycle with the exception in self._exception

TerminatedWorkerError: A worker process managed by the executor was unexpectedly terminated. This could be caused by a segmentation fault while calling the function or by an excessive memory usage causing the Operating System to kill the worker.

The exit codes of the workers are {SIGSEGV(-11)}

Versions

System:
    python: 3.8.12 | packaged by conda-forge | (default, Sep 16 2021, 01:38:21)  [Clang 11.1.0 ]
executable: /Users/glemaitre/mambaforge/envs/dev/bin/python
   machine: macOS-11.6-arm64-arm-64bit

Python dependencies:
          pip: 21.2.4
   setuptools: 58.2.0
      sklearn: 1.1.dev0
        numpy: 1.21.2
        scipy: 1.7.1
       Cython: 0.29.24
       pandas: 1.3.3
   matplotlib: 3.4.3
       joblib: 1.0.1
threadpoolctl: 3.0.0

Built with OpenMP: True

[glemaitre]

Tomorrow is Friday. It could be a nide day to release 😃

On Thu, 14 Oct 2021 at 19:49, Olivier Grisel @.***> wrote:

We should probably hurry the 1.0.1 release for this and for #21188 https://github.com/scikit-learn/scikit-learn/issues/21188.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/scikit-learn/scikit-learn/issues/21283#issuecomment-943583794, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABY32P42XEUSAUK3AHT63ZLUG4J35ANCNFSM5FTTNRCA .

– Guillaume Lemaitre Scikit-learn @ Inria Foundation https://glemaitre.github.io/

[ogrisel]

I think we can consider that #21227 will fix it in 1.0.1.