Numpy: Optimal way to count indexs occurrence in an array

Question

I have an array indexs. It\'s very long (>10k), and each int value is rather small (<100). e.g.

inde