Converted XGBoost model outputs wrong results

[meakbiyik]

Hello, I found a parameter set where the converted XGBoost model is inconsistent with the original. Here's the minimal reproduction:

from onnxmltools.convert import convert_xgboost
import onnxruntime as rt
import numpy as np
from xgboost import XGBRegressor

from skl2onnx.common.data_types import (
    FloatTensorType,
)

df = pd.DataFrame(
    {
        "f1": [1.0, 2.0, 3.0, 4.0, 2.0, 3.0, 1.0, 2.0],
        "label": [1, 0, 1, 0, 1, 1, 0, 1],
    }
)
params = {
    "max_depth": 1,
    "n_estimators": 3,
    "subsample": 0.95,
    "objective": "binary:logistic",
}

model = XGBRegressor(**params)

initial_types = [
    ("f1", FloatTensorType([None, 1])),
]
model.fit(df.drop(columns=["label"]), df["label"])
onnx_model = convert_xgboost(
    model,
    "XGBoostXGBRegressor",
    initial_types,
    target_opset=13,
)
assert onnx_model is not None and hasattr(onnx_model, "SerializeToString")

sess = rt.InferenceSession(onnx_model.SerializeToString(), providers=["CPUExecutionProvider"])
onnx_output = sess.run(
    None,
    {
        "f1": df["f1"].values.reshape(-1, 1).astype(np.float32),
    },
)[0]

expected_output = model.predict(df.drop(columns=["label"])).reshape(-1, 1).astype(np.float32)
assert np.allclose(
    onnx_output, expected_output, rtol=1e-5, atol=1e-8, equal_nan=True
), f"ONNX output does not match expected values for params: {params}"

Which outputs;

AssertionError: ONNX output does not match expected values for params: {'max_depth': 1, 'n_estimators': 3, 'subsample': 0.95, 'objective': 'binary:logistic'}
E       assert False
E        +  where False = <function allclose at 0x106fe80f0>(array([[0.69600594],\n       [0.69600594],\n       [0.69600594],\n       [0.69600594],\n       [0.69600594],\n       [0.69600594],\n       [0.69600594],\n       [0.69600594]], dtype=float32), array([[0.64148873],\n       [0.64148873],\n       [0.64148873],\n       [0.64148873],\n       [0.64148873],\n       [0.64148873],\n       [0.64148873],\n       [0.64148873]], dtype=float32), rtol=1e-05, atol=1e-08, equal_nan=True)

I noted that removing "subsample" parameter also fixes the issue, not sure what this means.

Versions:

xgboost==3.0.2 ; python_version >= "3.11" and python_version < "3.14"
onnx==1.18.0 ; python_version >= "3.11" and python_version < "3.14"
onnxmltools==1.14.0 ; python_version >= "3.11" and python_version < "3.14"
onnxruntime==1.22.0 ; python_version >= "3.11" and python_version < "3.14"