Question Required constant "version_number" was not found in the model file.

I train a model via pytorch and convert it into .onnx
But when I imported the .onnx to Unity
It shows Required constant "version_number" was not found in the model file.


Then I use WinML to see what I'm missing
I see this on the model that working on Unity



But my model is missing it


How could I add those outputs in my model?

Here's my code on pytorch side.

Code (Boo):

1. class Network(nn.Module):
2.     def __init__(self,version_number):
3.         super(Network, self).__init__()
4.        
5.         self.conv1 = nn.Conv2d(in_channels=3, out_channels=12, kernel_size=5, stride=1, padding=1)
6.         self.bn1 = nn.BatchNorm2d(12)
7.         self.conv2 = nn.Conv2d(in_channels=12, out_channels=12, kernel_size=5, stride=1, padding=1)
8.         self.bn2 = nn.BatchNorm2d(12)
9.         self.pool = nn.MaxPool2d(2,2)
10.         self.conv4 = nn.Conv2d(in_channels=12, out_channels=24, kernel_size=5, stride=1, padding=1)
11.         self.bn4 = nn.BatchNorm2d(24)
12.         self.conv5 = nn.Conv2d(in_channels=24, out_channels=24, kernel_size=5, stride=1, padding=1)
13.         self.bn5 = nn.BatchNorm2d(24)
14.         self.fc1 = nn.Linear(24*10*10, 10)
15.  
16.     def forward(self, input ):
17.         output = F.relu(self.bn1(self.conv1(input)))    
18.         output = F.relu(self.bn2(self.conv2(output)))    
19.         output = self.pool(output)                      
20.         output = F.relu(self.bn4(self.conv4(output)))    
21.         output = F.relu(self.bn5(self.conv5(output)))    
22.         output = output.view(-1, 24*10*10)
23.         output = self.fc1(output)
24.  
25.         return output
26.  
27. # Instantiate a neural network model
28. model = Network()
29.  
30. print(model)
31.  
32. # Define the loss function with Classification Cross-Entropy loss and an optimizer with Adam optimizer
33. loss_fn = nn.CrossEntropyLoss()
34. optimizer = Adam(model.parameters(), lr=0.001, weight_decay=0.0001)
35.  
36.     #Function to Convert to ONNX
37. def Convert_ONNX():
38.  
39.     # set the model to inference mode
40.     model.eval()
41.  
42.     # Let's create a dummy input tensor
43.     dummy_input = torch.randn(2, 3, 32, 32, requires_grad=True)
44.  
45.     inputNames = [ "actual_input_1" ]
46.     outputNames = [ "output1" ]
47.  
48.     # Export the model  
49.     torch.onnx.export(model,         # model being run
50.          dummy_input,       # model input (or a tuple for multiple inputs)
51.          "ImageClassifier.onnx",       # where to save the model
52.          export_params=True,  # store the trained parameter weights inside the model file
53.          opset_version=10,    # the ONNX version to export the model to
54.          do_constant_folding=True,  # whether to execute constant folding for optimization
55.          input_names = inputNames,   # the model's input names
56.          output_names = outputNames # the model's output names
57.          )
58.     print(" ")
59.     print('Model has been converted to ONNX')
60.  
61.  

 

I have the same question.Please tell me if someone knows the answer.Thanks!