[FIXED] pytorch inplace runtime error with class inheritance

Issue

May I know why this forward() function gives runtime error on inplace operation ?

Note: I have done some code debugging which leads to the following line of code:

class ConvEdge(Edge):    
    def __init__(self, stride): 
        super().__init__()        
        self.f = nn.Conv2d(in_channels=3, out_channels=3, kernel_size=(3, 3), stride=(stride, stride), padding=1)

If you guys have a look at the code snippet regarding class ConvEdge(Edge) , I am actually having second thought on how inheritance is being viewed and processed by pytorch autograd library.

What do you guys think ?

Solution

the issue is resolved using with torch.no_grad() which basically does not propagate the gradient when it is not necessary.

Answered By - kevin