for example,in pytorch version 0.2,
modle = nn.Conv2d(in_channels=3 out_channels=32, kernel_size=3,stride=2, padding=1, dilation=1, bias=False))
y = modle(x)
if I get the result y, I can get the input param’s values like kernel_size、stride、padding value use y.grad_fn.kernel_size,y.grad_fn.padding,……
but in pytorch version 0.3,the grad_fn(which has type ‘thnnconv2dbackward’) has not attribute like kernel_size, padding……
So,how could I get these input params’s values in the newest version of pytorch?