The easiest way to scale down would be to take the first 100 iterations of t but this may not be very salient so you might want pick certain iterations first and spread them out.