3035 enhance ModelWithHooks error handling

monai/visualize/class_activation_maps.py

@@ -137,6 +137,11 @@ def __call__(self, x, class_idx=None, retain_graph=False):
             self.score = self.class_score(logits, self.class_idx)
             self.model.zero_grad()
             self.score.sum().backward(retain_graph=retain_graph)
+            for layer in self.target_layers:
+                if layer not in self.gradients:
+                    raise RuntimeError(
+                        f"Backward hook for {layer} is not triggered; `requires_grad` of {layer} should be `True`."
+                    )
             grad = tuple(self.gradients[layer] for layer in self.target_layers)
         if train:
             self.model.train()
@@ -221,6 +226,8 @@ class CAM(CAMBase):
 
     .. code-block:: python
 
+        import torch
+
         # densenet 2d
         from monai.networks.nets import DenseNet121
         from monai.visualize import CAM
@@ -319,6 +326,8 @@ class GradCAM(CAMBase):
 
     .. code-block:: python
 
+        import torch
+
         # densenet 2d
         from monai.networks.nets import DenseNet121
         from monai.visualize import GradCAM

tests/test_vis_gradcam.py

@@ -88,6 +88,16 @@ def test_shape(self, input_data, expected_shape):
         result2 = cam(x=image, layer_idx=-1, class_idx=model(image).max(1)[-1].cpu())
         torch.testing.assert_allclose(result, result2)
 
+    def test_ill(self):
+        model = DenseNet121(spatial_dims=2, in_channels=1, out_channels=3)
+        for name, x in model.named_parameters():
+            if "features" in name:
+                x.requires_grad = False
+        cam = GradCAM(nn_module=model, target_layers="class_layers.relu")
+        image = torch.rand((2, 1, 48, 64))
+        with self.assertRaises(RuntimeError):
+            cam(x=image)
+
 
 if __name__ == "__main__":
     unittest.main()