Rename to functional and add 3d conv test

Author: Alejandro Gaston Alvarez Franceschi

coremltools/converters/mil/frontend/torch/test/test_torch_ops.py

@@ -1640,7 +1640,7 @@ def test_convolution3d(
         )
 
 
-class TestDynamicConv(TorchBaseTest):
+class TestFunctionalConv(TorchBaseTest):
     @pytest.mark.parametrize(
         ",".join(
             [
@@ -1687,13 +1687,13 @@ def test_convolution1d(
         padding,
         groups=1,
     ):
-        class DynamicConv(nn.Module):
+        class FunctionalConv1D(nn.Module):
             def forward(self, input_data, weights):
                 return nn.functional.conv1d(
                     input_data, weights, stride=stride, padding=padding, groups=groups
                 )
 
-        model = DynamicConv()
+        model = FunctionalConv1D()
         input_shape = [
             (1, in_channels, width),
             (out_channels, int(in_channels / groups), kernel_size),
@@ -1754,13 +1754,13 @@ def test_convolution2d(
         padding,
         groups=1,
     ):
-        class DynamicConv(nn.Module):
+        class FunctionalConv2D(nn.Module):
             def forward(self, input_data, weights):
                 return nn.functional.conv2d(
                     input_data, weights, stride=stride, padding=padding, groups=groups
                 )
 
-        model = DynamicConv()
+        model = FunctionalConv2D()
 
         input_shape = [
             (1, in_channels, height, width),
@@ -1774,21 +1774,19 @@ def forward(self, input_data, weights):
             use_scripting=use_scripting
         )
 
-
-class TestConvTranspose(TorchBaseTest):
     @pytest.mark.parametrize(
         ",".join(
             [
                 "compute_unit",
                 "backend",
                 "use_scripting",
+                "height",
                 "width",
                 "in_channels",
                 "out_channels",
                 "kernel_size",
                 "stride",
                 "padding",
-                "dilation",
             ]
         ),
         [
@@ -1797,6 +1795,74 @@ class TestConvTranspose(TorchBaseTest):
                 compute_units,
                 backends,
                 [True, False],
+                [
+                    (5, 3, 2, 1, 1, 1, 2, 0),
+                    (3, 3, 1, 1, 1, 1, 2, 1),
+                    (4, 3, 3, 3, 3, 1, 2, 0),
+                    (7, 3, 4, 3, 3, 1, 3, 0),
+                    (5, 5, 3, 3, 3, 2, 1, 0),
+                    (3, 5, 1, 3, 3, 1, 3, 0),
+                    (3, 5, 4, 3, 3, 1, 3, 1),
+                    (7, 5, 6, 3, 3, 2, 3, 1),
+                ],
+            )
+        ],
+    )
+    def test_convolution3d(
+        self,
+        compute_unit,
+        backend,
+        use_scripting,
+        height,
+        width,
+        in_channels,
+        out_channels,
+        kernel_size,
+        stride,
+        padding,
+        groups=1,
+    ):
+        class FunctionalConv3D(nn.Module):
+            def forward(self, input_data, weights):
+                return nn.functional.conv3d(
+                    input_data, weights, stride=stride, padding=padding, groups=groups
+                )
+
+        model = FunctionalConv3D()
+
+        input_shape = [
+            (1, in_channels, height, width),
+            (out_channels, int(in_channels / groups), kernel_size, kernel_size),
+        ]
+        self.run_compare_torch(
+            input_shape,
+            model,
+            backend=backend,
+            compute_unit=compute_unit,
+            use_scripting=use_scripting
+        )
+
+
+class TestConvTranspose(TorchBaseTest):
+    @pytest.mark.parametrize(
+        ",".join(
+            [
+                "compute_unit",
+                "backend",
+                "width",
+                "in_channels",
+                "out_channels",
+                "kernel_size",
+                "stride",
+                "padding",
+                "dilation",
+            ]
+        ),
+        [
+            (compute_unit, backend, *param)
+            for compute_unit, backend, param in itertools.product(
+                compute_units,
+                backends,
                 [
                     (3, 1, 1, 1, 2, 0, 1),
                     (3, 1, 1, 1, 2, 1, 3),