pandas groupby() and memory consumption

Question

I am using following code to create and array the have indexes of unique entries. Since I need to use .compute() to use ngroup() it adds extra memo