Backpropagation with multiple loss functions

1

How can model backpropagate with a combination of multiple loss functions?
Example: loss = loss1 + loss2 + loss3

2

(loss1 + loss2 + loss3).backward()