understanding pytorch sample code for gradient calculation

Question

I do not understand the purpose of the following line of code:

external_grad = torch.tensor([1., 1.])
Q.backward(gradient=external_grad)

Here