Print shape of a tensor

autograd
1

I am a beginner in pytorch. I have the following neural network
class FirstNN(nn.Module):
def init(self):
super(FirstNN,self).init()
self.fc1 = nn.Linear(784,hidden_size)
self.act1 = nn.ReLU()
self.fc2 = nn.Linear(hidden_size,num_classes)

def forward(self,x):
    out = self.fc1(x)
    print(out.Size())
    out = self.act1(out)
    out = self.fc2(out)
    return out

model = FirstNN()

I wanted to know the shape of tensor after fc1. When I tried using print function it doesn’t print anything.How can I know the shape of tensor??

2

Did you pass a random sample through your model?

x = torch.randn(1, 784)
model = FirstNN()
out = model(x)
3

No I haven’t. It’s working now. Thanks a lot.

4

@avinash_m How did you print shape of tensor after fc1?

5

The print statement is inside the ‘forward’ function. However, you need to call the function to see the result. You call the function as given above:

x = torch.randn(1, 784)
model = FirstNN()
out = model(x) #This makes a call to forward