[CUDA] max_depth parameter is ignored #6969
Description
Description
When training a lightgbm model with device="cuda" the max_depth parameter appears to be ignored. When using instead device="cpu" or device="gpu" (i.e. OpenCL implementation) the max_depth constraint is enforced.
Reproducible example
import lightgbm as lgb
from sklearn.datasets import make_regression
# Generate synthetic regression data
X, y = make_regression(n_samples=1000, n_features=20, noise=0.1, random_state=42)
# Create and fit the LGBMRegressor with GPU support
model = lgb.LGBMRegressor(
objective="regression",
device="cuda", # Use CUDA
max_depth=10,
)
model.fit(X, y)
print(model.booster_.trees_to_dataframe().groupby("tree_index")["node_depth"].max())tree_index
0 8
1 9
2 9
3 9
4 9
..
95 18
96 14
97 18
98 17
99 14
Name: node_depth, Length: 100, dtype: int64
Environment info
LightGBM version or commit hash: 4.6.0
Command(s) you used to install LightGBM
pixi add lightgbm