When we derive energy momentum tensor current by actively transforming field. We see that lagrangian ( density) changes by a total derivative of the lagrangian. If a total derivative of the function is added to a lagrangian, i know the action will remain invariant becausr its variation will be zero because at the boundaries variation of field is zero. But in the energy momentum tensor case we have a total derivative of the lagrangian itself which is a function of fields AND THEIR DERIVATIVES, why should action still remain invariant? I mean the variation of surface term is not zero in this case because we have derivatives of fields whose variations on the boundary is not necessarily zero?