Fix nn functional conv2d bug

oneflow/core/functional/functional_api.yaml

@@ -696,23 +696,23 @@
 
 - name: "conv1d"
   signature:
-    "Tensor (Tensor x, Tensor weight, Tensor bias=None, Int32List[1] stride,
-    Int32List[1] padding, Int32List[1] dilation, Int32 groups=1,
-    String channel_pos) => Conv1d"
+    "Tensor (Tensor x, Tensor weight, Tensor bias=None, Int32List[1] stride={1},
+    Int32List[1] padding={0}, Int32List[1] dilation={1}, Int32 groups=1,
+    String channel_pos=\"channels_first\") => Conv1d"
   bind_python: True
 
 - name: "conv2d"
   signature:
-    "Tensor (Tensor x, Tensor weight, Tensor bias=None, Int32List[2] stride,
-    Int32List[2] padding, Int32List[2] dilation, Int32 groups=1,
-    String channel_pos) => Conv2d"
+    "Tensor (Tensor x, Tensor weight, Tensor bias=None, Int32List[2] stride={1},
+    Int32List[2] padding={0}, Int32List[2] dilation={1}, Int32 groups=1,
+    String channel_pos=\"channels_first\") => Conv2d"
   bind_python: True
 
 - name: "conv3d"
   signature:
-    "Tensor (Tensor x, Tensor weight, Tensor bias=None, Int32List[3] stride,
-    Int32List[3] padding, Int32List[3] dilation, Int32 groups=1,
-    String channel_pos) => Conv3d"
+    "Tensor (Tensor x, Tensor weight, Tensor bias=None, Int32List[3] stride={1},
+    Int32List[3] padding={0}, Int32List[3] dilation={1}, Int32 groups=1,
+    String channel_pos=\"channels_first\") => Conv3d"
   bind_python: True
 
 - name: "fake_quantization"

python/oneflow/test/modules/test_conv1d.py

@@ -435,6 +435,15 @@ def test_conv1d(test_case):
         for arg in GenArgList(arg_dict):
             arg[0](test_case, *arg[1:])
 
+    def test_nn_functional_conv1d(test_case):
+        img = flow.ones(1, 3, 224)
+        kernel = flow.randn(3, 1, 3)
+        y = flow.nn.functional.conv1d(img, kernel, groups=3)
+        assert y.shape[0] == img.shape[0]
+        assert y.shape[1] == kernel.shape[0]
+        # default padding is 1.
+        assert y.shape[2] == img.shape[2] - 2
+
     @autotest()
     def test_conv1d_with_random_data(test_case):
         channels = random(1, 6)

python/oneflow/test/modules/test_conv2d.py

@@ -1581,6 +1581,16 @@ def test_conv2d_default_init(test_case):
                 )
             )
 
+    def test_nn_functional_conv2d(test_case):
+        img = flow.ones(1, 3, 224, 224)
+        kernel = flow.randn(3, 1, 3, 3)
+        y = flow.nn.functional.conv2d(img, kernel, groups=3)
+        assert y.shape[0] == img.shape[0]
+        assert y.shape[1] == kernel.shape[0]
+        # default padding is 1.
+        assert y.shape[2] == img.shape[2] - 2
+        assert y.shape[3] == img.shape[3] - 2
+
     def test_conv2d(test_case):
         arg_dict = OrderedDict()
         arg_dict["device"] = ["cuda", "cpu"]

python/oneflow/test/modules/test_conv3d.py

@@ -22,6 +22,17 @@
 
 @flow.unittest.skip_unless_1n1d()
 class TestConv3DModule(flow.unittest.TestCase):
+    def test_nn_functional_conv3d(test_case):
+        img = flow.ones(1, 3, 224, 224, 224)
+        kernel = flow.randn(6, 3, 3, 3, 3)
+        y = flow.nn.functional.conv3d(img, kernel)
+        assert y.shape[0] == img.shape[0]
+        assert y.shape[1] == kernel.shape[0]
+        # default padding is 1.
+        assert y.shape[2] == img.shape[2] - 2
+        assert y.shape[3] == img.shape[3] - 2
+        assert y.shape[3] == img.shape[3] - 2
+
     @autotest(n=10)
     def test_conv3d_with_random_data(test_case):
         channels = random(1, 6)