Expected stride to be a single integer value or a list?

vision
21

So I went through this thread and I am still facing this error. In my case, I am using a 3rd party library to just see how the network is working (torchinfo). I am giving the input as (1,1,28,28). My model is as follows:

net(
  (conv1): Conv2d(1, 1, kernel_size=(2, 2), stride=(1, 1))
  (pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (conv2): Conv2d(1, 1, kernel_size=(2, 2), stride=(1, 1))
)

But, I get

RuntimeError: expected stride to be a single integer value or a list of 1 values to match the convolution dimensions, but got stride=[1, 1]

when I run torchinfo.summary(net, (1,1,28,28)).
Torchinfo seems to be running fine otherwise. What should I do?

22

Hello~

        self.pca1 = nn.Conv2d(
            in_channels=rff_num_filter1,
            out_channels=pca_num_filter1,
            kernel_size=10,
            padding=round((kernel_size1-1)/2)
        ).requires_grad_(False)

data : torch.Size([1, 100, 128, 128])
Conv2d(100, 8, kernel_size=(10, 10), stride=(1, 1), padding=(3, 3))
The above is the dimension of the data and Conv2d ,but i had a question.can you help me ,please?

RuntimeError: expected stride to be a single integer value or a list of 1 values to match the convolution dimensions, but got stride=[1, 1]

23

Your code works fine:

pca1 = nn.Conv2d(in_channels=100, out_channels=8, kernel_size=10, padding=3).requires_grad_(False)

data = torch.randn(1, 100, 128, 128)
out = pca1(data)
print(out.shape)
# torch.Size([1, 8, 125, 125])

Could you check what the difference between my working code and yours would be?

24

yes,thank your reply, it’s right! then ,how i to do?

25

As I cannot reproduce the issue, could you post a minimal, executable code snippet which would raise the error?
Your current code snippet is not executable and my code works fine.

26

Hello. My PCa1 structure output is conv2d (100, 8, kernel_size= (10, 10), stripe= (1, 1), padding= (3, 3)), is it the same as yours?

27

Hello. My PCa1 structure output is conv2d (100, 8, kernel_size= (10, 10), stripe= (1, 1), padding= (3, 3)), is it the same as yours?

28

When I execute your code in jupyter cell alone, it works normally. But when I copy your code into my project (including data generated randomly), he will report the above error.