Pytorch model production

Is it good time to use jit in production level?
or tensorflow serving is better?