Fix unexpected type for intercept in PoissonRegressor and GammaRegressor

CHANGELOGS.md

@@ -2,6 +2,8 @@
 
 ## 1.17.0 (development)
 
+* Fix unexpected type for intercept in PoissonRegressor and GammaRegressor
+  [#1070](https://github.com/onnx/sklearn-onnx/pull/1070)
 * Add support for scikti-learn 1.4.0,
   [#1058](https://github.com/onnx/sklearn-onnx/pull/1058),
   fixes issues [Many examples in the gallery are showing "broken"](https://github.com/onnx/sklearn-onnx/pull/1057),

pyproject.toml

@@ -10,8 +10,8 @@ exclude = [
 # Same as Black.
 line-length = 88
 
-[tool.ruff.mccabe]
+[tool.ruff.lint.mccabe]
 max-complexity = 10
 
-[tool.ruff.per-file-ignores]
+[tool.ruff.lint.per-file-ignores]
 "skl2onnx/algebra/onnx_ops.py" = ["F821"]

skl2onnx/operator_converters/gamma_regressor.py

@@ -34,7 +34,7 @@ def convert_sklearn_gamma_regressor(
 
     intercept = (
         op.intercept_.astype(dtype)
-        if len(op.intercept_.shape) > 0
+        if isinstance(op.intercept_, np.ndarray) and len(op.intercept_.shape) > 0
         else np.array([op.intercept_], dtype=dtype)
     )
     eta = OnnxAdd(

skl2onnx/operator_converters/linear_regressor.py

@@ -198,7 +198,7 @@ def convert_sklearn_poisson_regressor(
 
     intercept = (
         op.intercept_.astype(dtype)
-        if len(op.intercept_.shape) > 0
+        if isinstance(op.intercept_, np.ndarray) and len(op.intercept_.shape) > 0
         else np.array([op.intercept_], dtype=dtype)
     )
     eta = OnnxAdd(

tests/test_sklearn_gamma_regressor.py

@@ -6,7 +6,7 @@
 import numpy as np
 
 try:
-    from sklearn.linear_model import GammaRegressor
+    from sklearn.linear_model import GammaRegressor, PoissonRegressor
 except ImportError:
     GammaRegressor = None
 from onnxruntime import __version__ as ort_version
@@ -90,6 +90,38 @@ def test_gamma_regressor_double(self):
             basename="SklearnGammaRegressor",
         )
 
+    @unittest.skipIf(GammaRegressor is None, reason="scikit-learn<1.0")
+    def test_poisson_without_intercept(self):
+        # Poisson
+        model = PoissonRegressor(fit_intercept=False)
+        X = np.array([[1.0, 2.0], [2.0, 3.0], [3.0, 4.0], [4.0, 3.0]])
+        y = np.array([19.0, 26.0, 33.0, 30.0])
+        model.fit(X, y)
+
+        model_onnx = convert_sklearn(
+            model,
+            "scikit-learn Poisson Regressor without Intercept",
+            [("input", FloatTensorType([None, X.shape[1]]))],
+        )
+
+        self.assertIsNotNone(model_onnx is not None)
+
+    @unittest.skipIf(GammaRegressor is None, reason="scikit-learn<1.0")
+    def test_gamma_without_intercept(self):
+        # Gamma
+        model = GammaRegressor(fit_intercept=False)
+        X = np.array([[1.0, 2.0], [2.0, 3.0], [3.0, 4.0], [4.0, 3.0]])
+        y = np.array([19.0, 26.0, 33.0, 30.0])
+        model.fit(X, y)
+
+        model_onnx = convert_sklearn(
+            model,
+            "scikit-learn Gamma Regressor without Intercept",
+            [("input", FloatTensorType([None, X.shape[1]]))],
+        )
+
+        self.assertIsNotNone(model_onnx is not None)
+
 
 if __name__ == "__main__":
     unittest.main(verbosity=3)