forcedsplits does not check features index boundary

[elena-sharova]

Description

When using forcedsplits_filename parameter, there seems to be no check on the boundaries of the provided features index.
This makes it possible to get an access violation error. Ideally there should be a check to avoid run-time crashes.

Reproducible example

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
import pandas as pd
import json
from lightgbm import LGBMClassifier

df = load_iris(return_X_y=True, as_frame=True)[0]
df.columns = [c.replace(" ","_") for c in df.columns]
x_cols = df.columns
df['target']=load_iris(return_X_y=True, as_frame=True)[1]

# The dataset has four features
X_train, X_test, y_train, y_test = train_test_split(df[x_cols], df['target'], test_size=0.33, random_state=42)

lgbm_model = LGBMClassifier(random_state=42)

# we proceed as if the dataset has five features
spl_dict = {"feature": 4,
 "threshold": 3.0,
 "left": {
        "feature": 0,
        "threshold": 3.25
    },
"right": {
    "feature": 0,
    "threshold": 3.25
    }
}

# write dict to a json file
with open('splits.json', 'w') as fp:
    json.dump(spl_dict, fp)

lgbm_model = LGBMClassifier(random_state=42,forcedsplits_filename="splits.json")
lgbm_model.fit(X_train, y_train)

The above gives this error:

OSError                                   Traceback (most recent call last)
Input In [39], in <cell line: 1>()
----> 1 lgbm_model.fit(X_train, y_train)

File ~\AppData\Local\Continuum\anaconda3\envs\ml_env\lib\site-packages\lightgbm\sklearn.py:967, in LGBMClassifier.fit(self, X, y, sample_weight, init_score, eval_set, eval_names, eval_sample_weight, eval_class_weight, eval_init_score, eval_metric, early_stopping_rounds, verbose, feature_name, categorical_feature, callbacks, init_model)
    964         else:
    965             valid_sets[i] = (valid_x, self._le.transform(valid_y))
--> 967 super().fit(X, _y, sample_weight=sample_weight, init_score=init_score, eval_set=valid_sets,
    968             eval_names=eval_names, eval_sample_weight=eval_sample_weight,
    969             eval_class_weight=eval_class_weight, eval_init_score=eval_init_score,
    970             eval_metric=eval_metric, early_stopping_rounds=early_stopping_rounds,
    971             verbose=verbose, feature_name=feature_name, categorical_feature=categorical_feature,
    972             callbacks=callbacks, init_model=init_model)
    973 return self

File ~\AppData\Local\Continuum\anaconda3\envs\ml_env\lib\site-packages\lightgbm\sklearn.py:748, in LGBMModel.fit(self, X, y, sample_weight, init_score, group, eval_set, eval_names, eval_sample_weight, eval_class_weight, eval_init_score, eval_group, eval_metric, early_stopping_rounds, verbose, feature_name, categorical_feature, callbacks, init_model)
    745 evals_result = {}
    746 callbacks.append(record_evaluation(evals_result))
--> 748 self._Booster = train(
    749     params=params,
    750     train_set=train_set,
    751     num_boost_round=self.n_estimators,
    752     valid_sets=valid_sets,
    753     valid_names=eval_names,
    754     fobj=self._fobj,
    755     feval=eval_metrics_callable,
    756     init_model=init_model,
    757     feature_name=feature_name,
    758     callbacks=callbacks
    759 )
    761 if evals_result:
    762     self._evals_result = evals_result

File ~\AppData\Local\Continuum\anaconda3\envs\ml_env\lib\site-packages\lightgbm\engine.py:292, in train(params, train_set, num_boost_round, valid_sets, valid_names, fobj, feval, init_model, feature_name, categorical_feature, early_stopping_rounds, evals_result, verbose_eval, learning_rates, keep_training_booster, callbacks)
    284 for cb in callbacks_before_iter:
    285     cb(callback.CallbackEnv(model=booster,
    286                             params=params,
    287                             iteration=i,
    288                             begin_iteration=init_iteration,
    289                             end_iteration=init_iteration + num_boost_round,
    290                             evaluation_result_list=None))
--> 292 booster.update(fobj=fobj)
    294 evaluation_result_list = []
    295 # check evaluation result.

File ~\AppData\Local\Continuum\anaconda3\envs\ml_env\lib\site-packages\lightgbm\basic.py:3021, in Booster.update(self, train_set, fobj)
   3019 if self.__set_objective_to_none:
   3020     raise LightGBMError('Cannot update due to null objective function.')
-> 3021 _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
   3022     self.handle,
   3023     ctypes.byref(is_finished)))
   3024 self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
   3025 return is_finished.value == 1

OSError: exception: access violation reading 0x0000025A797419E8

Environment info

LightGBM version or commit hash: 3.3.2

Command(s) you used to install LightGBM: conda install -c conda-forge lightgbm

Additional Comments

[btrotta]

@elena-sharova Thanks for reporting this, and for the reproducible example. I've opened a PR to fix this.

[jameslamb]

+1 to that, excellent bug report @elena-sharova . Thank you so much for the effort you put into it.

[github-actions]

This issue has been automatically locked since there has not been any recent activity since it was closed. To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues including a reference to this.