Fix for AnchorGenerator when device switch happen (#1745)

* Fix AnchorGenerator if moving from one device to another

* Fixes for the test

test/test_models.py

@@ -226,6 +226,28 @@ def test_fasterrcnn_double(self):
         self.assertTrue("scores" in out[0])
         self.assertTrue("labels" in out[0])
 
+    @unittest.skipIf(not torch.cuda.is_available(), 'needs GPU')
+    def test_fasterrcnn_switch_devices(self):
+        model = models.detection.fasterrcnn_resnet50_fpn(num_classes=50, pretrained_backbone=False)
+        model.cuda()
+        model.eval()
+        input_shape = (3, 300, 300)
+        x = torch.rand(input_shape, device='cuda')
+        model_input = [x]
+        out = model(model_input)
+        self.assertIs(model_input[0], x)
+        self.assertEqual(len(out), 1)
+        self.assertTrue("boxes" in out[0])
+        self.assertTrue("scores" in out[0])
+        self.assertTrue("labels" in out[0])
+        # now switch to cpu and make sure it works
+        model.cpu()
+        x = x.cpu()
+        out_cpu = model([x])
+        self.assertTrue("boxes" in out_cpu[0])
+        self.assertTrue("scores" in out_cpu[0])
+        self.assertTrue("labels" in out_cpu[0])
+
 
 for model_name in get_available_classification_models():
     # for-loop bodies don't define scopes, so we have to save the variables

torchvision/models/detection/rpn.py

@@ -90,7 +90,12 @@ def generate_anchors(self, scales, aspect_ratios, dtype=torch.float32, device="c
     def set_cell_anchors(self, dtype, device):
         # type: (int, Device) -> None    # noqa: F821
         if self.cell_anchors is not None:
-            return
+            cell_anchors = self.cell_anchors
+            assert cell_anchors is not None
+            # suppose that all anchors have the same device
+            # which is a valid assumption in the current state of the codebase
+            if cell_anchors[0].device == device:
+                return
 
         cell_anchors = [
             self.generate_anchors(