Dynamic ResNet (#2474)

Summary:
Pull Request resolved: #2474

Test for dynamic resnet.

ResNet has some restrictions on the input shape, so we create a dynamic version by bilinear resizing the input to resnet's fixed shape. Thus we test that dynamic bilinear resize correctly resizes to fixed shape

Reviewed By: digantdesai

Differential Revision: D54972682

Author: mcr229

backends/xnnpack/test/models/resnet.py

@@ -14,27 +14,63 @@
 
 
 class TestResNet18(unittest.TestCase):
-    def test_fp32_resnet18(self):
-        inputs = (torch.ones(1, 3, 224, 224),)
+    inputs = (torch.ones(1, 3, 224, 224),)
+    dynamic_shapes = (
+        {
+            2: torch.export.Dim("height", min=224, max=455),
+            3: torch.export.Dim("width", min=224, max=455),
+        },
+    )
+
+    class DynamicResNet(torch.nn.Module):
+        def __init__(self):
+            super().__init__()
+            self.model = torchvision.models.resnet18()
+
+        def forward(self, x):
+            x = torch.nn.functional.interpolate(
+                x,
+                size=(224, 224),
+                mode="bilinear",
+                align_corners=True,
+                antialias=False,
+            )
+            return self.model(x)
+
+    def _test_exported_resnet(self, tester):
         (
-            Tester(torchvision.models.resnet18(), inputs)
-            .export()
+            tester.export()
             .to_edge()
             .partition()
+            .check_not(
+                [
+                    "executorch_exir_dialects_edge__ops_aten_convolution_default",
+                    "executorch_exir_dialects_edge__ops_aten_mean_dim",
+                ]
+            )
+            .check(["torch.ops.higher_order.executorch_call_delegate"])
             .to_executorch()
             .serialize()
             .run_method_and_compare_outputs()
         )
 
+    def test_fp32_resnet18(self):
+        self._test_exported_resnet(Tester(torchvision.models.resnet18(), self.inputs))
+
     def test_qs8_resnet18(self):
-        inputs = (torch.ones(1, 3, 224, 224),)
-        (
-            Tester(torchvision.models.resnet18(), inputs)
-            .quantize(Quantize(calibrate=False))
-            .export()
-            .to_edge()
-            .partition()
-            .to_executorch()
-            .serialize()
-            .run_method_and_compare_outputs()
+        quantized_tester = Tester(torchvision.models.resnet18(), self.inputs).quantize(
+            Quantize(calibrate=False)
+        )
+        self._test_exported_resnet(quantized_tester)
+
+    def test_fp32_resnet18_dynamic(self):
+        self._test_exported_resnet(
+            Tester(self.DynamicResNet(), self.inputs, self.dynamic_shapes)
+        )
+
+    def test_qs8_resnet18_dynamic(self):
+        self._test_exported_resnet(
+            Tester(self.DynamicResNet(), self.inputs, self.dynamic_shapes).quantize(
+                Quantize(calibrate=False)
+            )
         )