Exclude combined_non_maximum_supression from Hessian compatible nodes

tests/keras_tests/function_tests/test_sensitivity_eval_non_suppoerted_output.py

@@ -38,6 +38,33 @@ def argmax_output_model(input_shape):
     model = keras.Model(inputs=inputs, outputs=outputs)
     return model
 
+def nms_output_model(input_shape):
+    inputs = layers.Input(shape=input_shape)
+    x = layers.Conv2D(1, 3, padding='same')(inputs)
+    x = layers.ReLU()(x)
+
+    # Dummy layers for creating NMS inputs with the required shape
+    x = tf.squeeze(x, -1)
+    x = tf.concat([x, x], -1)
+    y = tf.concat([x, x], -1)
+    y = tf.concat([y, y], -1)
+    scores = tf.concat([x, y], -1) # shape = (batch, detections, classes)
+    boxes, _ = tf.split(x, (4,12), -1)
+    boxes = tf.expand_dims(boxes, 2) # shape = (batch, detections, 1, box coordinates)
+
+    # NMS layer
+    outputs = tf.image.combined_non_max_suppression(
+        boxes,
+        scores,
+        max_output_size_per_class=300,
+        max_total_size=300,
+        iou_threshold=0.7,
+        score_threshold=0.001,
+        pad_per_class=False,
+        clip_boxes=False
+    )
+    model = keras.Model(inputs=inputs, outputs=outputs)
+    return model
 
 def representative_dataset():
     yield [np.random.randn(1, 8, 8, 3).astype(np.float32)]
@@ -71,6 +98,11 @@ def test_not_supported_output_argmax(self):
             self.verify_test_for_model(model)
         self.assertTrue("All graph outputs should support Hessian computation" in str(e.exception))
 
+    def test_not_supported_output_nms(self):
+        model = nms_output_model((8, 8, 3))
+        with self.assertRaises(Exception) as e:
+            self.verify_test_for_model(model)
+        self.assertTrue("All graph outputs should support Hessian computation" in str(e.exception))
 
 if __name__ == '__main__':
     unittest.main()