When to use KLDivLoss and MSELoss?

Suppose the output is a continuous value in [0,1]. Is there any case we should prefer one or the other?