ver 4.0.0: lgb.cv returns a dict whose keys have an extra "valid" prefix

[xinwang0312]

Description

For the code below, lgb.cv version 3.3.2 gives a dictionary with keys
['binary_logloss-mean', 'binary_logloss-stdv']

but lgb version 4.0.0 gives an output
['valid binary_logloss-mean', 'valid binary_logloss-stdv']

Reproducible example

import lightgbm as lgb
from sklearn.datasets import load_iris
from sklearn.model_selection import KFold



X, y = load_iris(return_X_y=True, as_frame=True)
y = y > 0

data = lgb.Dataset(data=X, label=y)
metric={'name': 'binary_logloss', 'maximize': False}
folds=KFold(n_splits=3, shuffle=True, random_state=23)

params = {'boosting_type': 'gbdt', 'objective': 'binary', 
'verbose': -1, 'feature_pre_filter': False, 'num_leaves': 82, 'max_depth': 8, 
'learning_rate': 0.15543735428541233, 'min_data_in_leaf': 28, 'lambda_l2': 0.04420907265233052}

eval_hist = lgb.cv(
    params, train_set=data, folds=folds, seed=23, metrics='binary_logloss'
)
print(lgb.__version__)
print(eval_hist.keys())

Environment info

4.0.0

Command(s) you used to install LightGBM

conda install -c conda-forge lightgbm

Additional Comments

Screenshot of the result comparsion:

[jmoralez]

Hey @xinwang0312, thanks for using LightGBM. This was a breaking changed introduced in #4947 to be consistent with the names of the metrics, previously it would only have the prefix if you set eval_train_metric=True, now it always does. Sorry for the troubles.

[xinwang0312]

Hi @jmoralez, thanks for the quick reply. I was not aware of it. Will adapt my code to this change.

[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.