Spark accumulator used in transformation in ML lib

Question

I have read that Spark accumulators should be used only in actions and not in transformations since Spark can only guarantee that the add method will be called once when it