PyTorch CNN linear layer shape after conv2d

Question

I was trying to learn PyTorch and came across a tutorial where a CNN is defined like below,

class Net(Module):   
    def __init__(self):
        super(Net, self)