Confusion about n_iter

[Rhyssiyan]

Hi, thanks for your great job.
I have some confusion about the benchmark protocol. I find that n_iter is set to 1 by default in the code. On the other hand, in Sec 4. Exp of the paper, it says that "we use a fixed batch size of 10 samples and perform few iterations over a batch(1-5)". Does it mean that the final result is reported by selecting the best result from n_iter=1 to n_iter=5? Or take the average result from n_iter=1 to n_iter=5?

Thanks in advance.