How to generate sound timestamps from a song?

Question

I want to build a model which predicts timestamps for when to play, so I end up with a replica of a song using only these sounds.

This is my code:

import t