Add unit test to investigate issue 1046

Signed-off-by: xadupre <xadupre@microsoft.com>

tests/test_issues_2024.py

@@ -271,6 +271,138 @@ def Classifier(features: list[str]) -> base.BaseEstimator:
                 )
                 assert modelengine is not None
 
+    @ignore_warnings(category=(ConvergenceWarning, FutureWarning))
+    def test_issue_1046(self):
+        import pandas as pd
+        import numpy as np
+
+        from xgboost import XGBClassifier
+
+        from skl2onnx.common.data_types import (
+            FloatTensorType,
+            StringTensorType,
+            Int64TensorType,
+            DoubleTensorType,
+        )
+        from skl2onnx import convert_sklearn, to_onnx, update_registered_converter
+        from skl2onnx.common.shape_calculator import (
+            calculate_linear_classifier_output_shapes,
+        )
+
+        from onnxmltools.convert.xgboost.operator_converters.XGBoost import (
+            convert_xgboost,
+        )
+
+        from sklearn.impute import SimpleImputer
+        from sklearn.preprocessing import StandardScaler, OneHotEncoder
+        from sklearn.compose import ColumnTransformer
+        from sklearn.pipeline import Pipeline
+
+        def convert_dataframe_schema(df, drop=None):
+            inputs = []
+            for k, v in zip(df.columns, df.dtypes):
+                if drop is not None and k in drop:
+                    continue
+                if v == "int64":
+                    t = Int64TensorType([None, 1])
+                elif v == "float32":
+                    t = FloatTensorType([None, 1])
+                elif v == "float64":
+                    t = DoubleTensorType([None, 1])
+                else:
+                    t = StringTensorType([None, 1])
+                inputs.append((k, t))
+            return inputs
+
+        def get_categorical_features(df: pd.DataFrame):
+            dtype_ser = df.dtypes
+            categorical_features = dtype_ser[dtype_ser == "object"].index.tolist()
+            return categorical_features
+
+        numeric_transformer = Pipeline(
+            steps=[
+                ("imputer", SimpleImputer(strategy="median")),
+                ("scaler", StandardScaler()),
+            ]
+        )
+
+        categorical_transformer = Pipeline(
+            steps=[("onehot", OneHotEncoder(handle_unknown="ignore"))]
+        )
+
+        X_train = pd.DataFrame(
+            [
+                {"x1": 0.5, "c1": "A", "c2": "B"},
+                {"x1": 0.6, "c1": "B", "c2": "B"},
+                {"x1": 0.7, "c1": "A", "c2": "C"},
+                {"x1": 0.8, "c1": "A", "c2": "B"},
+                {"x1": 0.9, "c1": "B", "c2": "B"},
+                {"x1": 1.1, "c1": "A", "c2": "C"},
+                {"x1": 1.2, "c1": "B", "c2": "B"},
+                {"x1": 1.3, "c1": "A", "c2": "B"},
+            ]
+        )
+        y_train = (np.arange(X_train.shape[0]) % 3) % 2
+        X_train["x1"] = X_train["x1"].astype(np.float32)
+
+        categorical_features = get_categorical_features(X_train)
+        numeric_features = [f for f in X_train.columns if not f in categorical_features]
+
+        preprocessor = ColumnTransformer(
+            transformers=[
+                ("num", numeric_transformer, numeric_features),
+                ("cat", categorical_transformer, categorical_features),
+            ]
+        )
+
+        est_val = Pipeline(
+            steps=[
+                ("preprocessor", preprocessor),
+                (
+                    "classifier",
+                    XGBClassifier(
+                        enable_categorical=False,
+                        random_state=42,
+                    ),
+                ),
+            ]
+        )
+
+        est_val.fit(X_train, y_train)
+
+        # guess model input schema from training data
+        schema = convert_dataframe_schema(X_train)
+
+        # register XGBoost model converter
+        update_registered_converter(
+            XGBClassifier,
+            "XGBoostXGBClassifier",
+            calculate_linear_classifier_output_shapes,
+            convert_xgboost,
+            options={"nocl": [True, False], "zipmap": [True, False, "columns"]},
+        )
+
+        # convert sklearn pipeline to ONNX
+        model_onnx = convert_sklearn(
+            est_val,
+            "xgb_pipeline",
+            schema,
+            target_opset={"": 18, "ai.onnx.ml": 3},
+            options={"zipmap": False},
+        )
+
+        import onnxruntime
+
+        sess = onnxruntime.InferenceSession(
+            model_onnx.SerializeToString(),
+            providers=["CPUExecutionProvider"],
+        )
+
+        expected = est_val.predict_proba(X_train)
+        feeds = {c: X_train[c].values.reshape((-1, 1)) for c in X_train.columns}
+        got = sess.run(None, feeds)
+        np.testing.assert_allclose(expected, got[1], atol=1e-5)
+
 
 if __name__ == "__main__":
     unittest.main(verbosity=2)