Check failed: (best_split_info.left_count) > (0) when using the ''cegb_penalty_feature_coupled'' parameter

[ZhangTP1996]

Description

[LightGBM] [Fatal] Check failed: (best_split_info.left_count) > (0) at /__w/1/s/python-package/compile/src/treelearner/serial_tree_learner.cpp, line 653 .

Reproducible example

The error happens when I use the "cegb_penalty_feature_coupled" parameter. If we remove the parameter, the error disappears.
A minimal example to reproduce the error:

import lightgbm as lgb
import pandas as pd

X = pd.read_csv('./cegb_X.csv', index_col=0)
y = pd.read_csv('./cegb_Y.csv', index_col=0)
params = {"n_estimators": 1000, "importance_type": "gain", "num_leaves": 64,
          "cegb_penalty_feature_coupled": [1 for _ in range(X.shape[1])],
          "seed": 1, "n_jobs": 40}
gbm = lgb.LGBMRegressor(**params)
gbm.fit(X, y)

cegb_X.csv
cegb_Y.csv

[ZhangTP1996]

The problem is fixed by installing the latest version 3.3.2.99

[jameslamb]

Thanks very much for coming back to close this!

I suspect the issue you faced was the same as #3679, which was fixed in #5164.

[jameslamb]

And thank you very much for providing a reproducible example, it's greatly appreciated 😊

[ZhangTP1996]

And thank you very much for providing a reproducible example, it's greatly appreciated 😊

You are welcome 😊

[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.