Can someone explain each output size calculation for this generator?

class Generator$($nn$.$Module$)$:

def $\_\_$init$\_\_($self$,$ z$\_$dim$=10,$ im$\_$chan$=1,$ hidden$\_$dim$=64)$:

super$($Generator$,$ self$).\_\_$init$\_\_()$

self.z$\_$dim $=$ z$\_$dim

# Build the neural network

self.gen $=$ nn$.$Sequential$($

self.make$\_$gen$\_$block$($z$\_$dim, hidden$\_$dim $*4),$

self.make$\_$gen$\_$block$($hidden$\_$dim $*4,$ hidden$\_$dim $*2,$ kernel$\_$size$=4,$ stride$=1),$

self.make$\_$gen$\_$block$($hidden$\_$dim $*2,$ hidden$\_$dim$),$

self.make$\_$gen$\_$block$($hidden$\_$dim, im$\_$chan, kernel$\_$size$=4,$ stride$=2,$ padding$=0,$ final$\_$layer$=$True$),$

$)$

def make$\_$gen$\_$block$($self$,$ input$\_$channels, output$\_$channels, kernel$\_$size$=3,$ stride$=2,$ padding$=0,$final$\_$layer$=$False$)$:

layers $=[]$

layers.append$($nn$.$ConvTranspose$2$d$($input$\_$channels, output$\_$channels, kernel$\_$size, stride, padding, output$\_$padding$=$padding$))$

if not final$\_$layer:

layers.append$($nn$.$BatchNorm$2$d$($output$\_$channels$))$

layers.append$($nn$.$ReLU$($True$))$

else:

layers.append$($nn$.$Tanh$())$

return nn$.$Sequential$(*$layers$)$

def unsqueeze$\_$noise$($self$,$ noise$)$:

return noise.view$($len$($noise$),$ self.z$\_$dim, $1,1)$

def forward$($self$,$ noise$)$:

x $=$ self.unsqueeze$\_$noise$($noise$)$

return self.gen$($x$)$

def get$\_$noise$($n$\_$samples, z$\_$dim, device$=$'cpu'$)$:

return torch.randn$($n$\_$samples, z$\_$dim, device$=$device$)$