Ref Counting vs backward(retain_graph=True)?

autograd
1

Hi,

Assume below is my code:

x = Variable(torch.ones(2, 2), requires_grad=True)
y = x + 2
z = y * y * 3
out = z.mean()

out.backward(retain_graph=True)

x,y,z,out = 0,0,0,0

Will the graph be freed up (deleted) at this point?

Thanks

2

yes it will be freed.