Adding testvector onnx files and the .py files that created them.

utests/models/conv2d_16_4x4x16_k1x3s1,1_g16_p0,0.py

+import os
+import sys
+import torch
+
+
+class Pad0Module(torch.nn.Module):
+    def __init__(self):
+        super(Pad0Module, self).__init__()
+        self.conv3x3_pad00 = torch.nn.Conv2d( in_channels=16, out_channels=16, kernel_size=(1,3), stride=(1,1), padding=(0,0), groups=16 )
+        
+    def forward(self, x):
+        y = self.conv3x3_pad00(x)
+        return y
+
+input_size = (1, 16, 4, 4)
+output_size = (1, 16, 4, 2)
+
+model = Pad0Module()
+input = torch.randn(*input_size)
+model.eval()
+output = model(input)
+
+
+# Output onnx model path
+onnx_model_path = os.path.join(os.path.dirname(sys.argv[0]), "conv2d_16_4x4x16_k1x3s1,1_g16_p0,0.onnx")
+# Convert to onnx
+torch.onnx.export(model, input, onnx_model_path, opset_version=10)

utests/models/conv2d_16_4x4x16_k3x1s1,1_g16_p0,0.py

+import os
+import sys
+import torch
+
+
+class Pad0Module(torch.nn.Module):
+    def __init__(self):
+        super(Pad0Module, self).__init__()
+        self.conv3x3_pad00 = torch.nn.Conv2d( in_channels=16, out_channels=16, kernel_size=(3,1), stride=(1,1), padding=(0,0), groups=16 )
+        
+    def forward(self, x):
+        y = self.conv3x3_pad00(x)
+        return y
+
+input_size = (1, 16, 4, 4)
+output_size = (1, 16, 2, 4)
+
+model = Pad0Module()
+input = torch.randn(*input_size)
+model.eval()
+output = model(input)
+
+
+# Output onnx model path
+onnx_model_path = os.path.join(os.path.dirname(sys.argv[0]), "conv2d_16_4x4x16_k3x1s1,1_g16_p0,0.onnx")
+# Convert to onnx
+torch.onnx.export(model, input, onnx_model_path, opset_version=10)

utests/models/conv2d_16_4x4x16_k3x3s1,1_g16_p0,0.py

+import os
+import sys
+import torch
+
+
+class Pad0Module(torch.nn.Module):
+    def __init__(self):
+        super(Pad0Module, self).__init__()
+        self.conv3x3_pad00 = torch.nn.Conv2d( in_channels=16, out_channels=16, kernel_size=(3,3), stride=(1,1), padding=(0,0), groups=16 )
+        
+    def forward(self, x):
+        y = self.conv3x3_pad00(x)
+        return y
+
+input_size = (1, 16, 4, 4)
+output_size = (1, 16, 2, 2)
+
+model = Pad0Module()
+input = torch.randn(*input_size)
+model.eval()
+output = model(input)
+
+
+# Output onnx model path
+onnx_model_path = os.path.join(os.path.dirname(sys.argv[0]), "conv2d_16_4x4x16_k3x3s1,1_g16_p0,0.onnx")
+# Convert to onnx
+torch.onnx.export(model, input, onnx_model_path, opset_version=10)

utests/models/conv2d_16_4x4x1_k1x3s1,1_g1_p0,0.py

+import os
+import sys
+import torch
+
+
+#  more7:  8 in 16 out <- pass
+#  more8: 16 in  8 out <- fail
+#  more9: 16 in  4 out <- fail
+# more10: 16 in  1 out <- fail
+# more11: 10x10 -> 10x8
+# more12: 10x10 -> 10x8
+
+class Pad0Module(torch.nn.Module):
+    def __init__(self):
+        super(Pad0Module, self).__init__()
+        self.conv1x3_pad00 = torch.nn.Conv2d( in_channels=16, out_channels=1, kernel_size=(1,3), stride=(1,1), padding=(0,0) )
+        
+    def forward(self, x):
+        y = self.conv1x3_pad00(x)
+        return y
+
+input_size = (1, 16, 4, 4)
+output_size = (1, 1, 4, 2)
+
+model = Pad0Module()
+input = torch.randn(*input_size)
+model.eval()
+output = model(input)
+
+
+# Output onnx model path
+onnx_model_path = os.path.join(os.path.dirname(sys.argv[0]), "conv2d_16_4x4x1_k1x3s1,1_g1_p0,0.onnx")
+# Convert to onnx
+torch.onnx.export(model, input, onnx_model_path, opset_version=10)

utests/models/conv2d_16_4x4x1_k3x1s1,1_g1_p0,0.py

+import os
+import sys
+import torch
+
+
+#  more7:  8 in 16 out <- pass
+#  more8: 16 in  8 out <- fail
+#  more9: 16 in  4 out <- fail
+# more10: 16 in  1 out <- fail
+# more11: 10x10 -> 10x8
+# more12: 10x10 -> 10x8
+
+class Pad0Module(torch.nn.Module):
+    def __init__(self):
+        super(Pad0Module, self).__init__()
+        self.conv1x3_pad00 = torch.nn.Conv2d( in_channels=16, out_channels=1, kernel_size=(3,1), stride=(1,1), padding=(0,0) )
+        
+    def forward(self, x):
+        y = self.conv1x3_pad00(x)
+        return y
+
+input_size = (1, 16, 4, 4)
+output_size = (1, 1, 2, 4)
+
+model = Pad0Module()
+input = torch.randn(*input_size)
+model.eval()
+output = model(input)
+
+
+# Output onnx model path
+onnx_model_path = os.path.join(os.path.dirname(sys.argv[0]), "conv2d_16_4x4x1_k3x1s1,1_g1_p0,0.onnx")
+# Convert to onnx
+torch.onnx.export(model, input, onnx_model_path, opset_version=10)

utests/models/conv2d_16_4x4x1_k3x3s1,1_g1_p0,0.py

+import os
+import sys
+import torch
+
+
+class Pad0Module(torch.nn.Module):
+    def __init__(self):
+        super(Pad0Module, self).__init__()
+        self.conv3x3_pad00 = torch.nn.Conv2d( in_channels=16, out_channels=1, kernel_size=(3,3), stride=(1,1), padding=(0,0) )
+        
+    def forward(self, x):
+        y = self.conv3x3_pad00(x)
+        return y
+
+input_size = (1, 16, 4, 4)
+output_size = (1, 1, 2, 2)
+
+model = Pad0Module()
+input = torch.randn(*input_size)
+model.eval()
+output = model(input)
+
+
+# Output onnx model path
+onnx_model_path = os.path.join(os.path.dirname(sys.argv[0]), "conv2d_16_4x4x1_k3x3s1,1_g1_p0,0.onnx")
+# Convert to onnx
+torch.onnx.export(model, input, onnx_model_path, opset_version=10)