Find audit specification with the following command (taking NSW 2015 election as an example) to generate an audit specification given a 5% risk limit and a 1% chance of an error being injected into each ballot. You can find some data files for use with this code in NSW2015.tar.gz (also available on github). 

./irvaudit -act_ballots Data_NA_Albury.txt_ballots.txt -rep_ballots Data_NA_Albury.txt_ballots.txt -r 0.05 -dive -s 39284575696785 -eseed 9284896769459 -eprob 0.01

You should get the following output:

440 errors added.
Reading Audits
Finished reading audits
============================================
AUDITS REQUIRED
EO,Winner,3,Loser,0,Eliminated,1,2,4
EO,Winner,0,Loser,1,Eliminated,2,3,4
EO,Winner,0,Loser,4,Eliminated,1,2,3
EO,Winner,0,Loser,2,Eliminated,1,3,4
WO,Winner,3,Loser,4,Eliminated
WO,Winner,3,Loser,2,Eliminated
EO,Winner,3,Loser,4,Eliminated,0,1,2
WO,Winner,3,Loser,1,Eliminated
EO,Winner,3,Loser,2,Eliminated,0,1,4
EO,Winner,3,Loser,1,Eliminated,0,2,4
MAX ASN(%) 0.20287
============================================
TIME,0.0971379,Nodes Expanded,0,MAX ASN(%),0.20287


This means that the algorithm found a series of facts (one per line) to audit, requiring an estimated 0.21% of total ballots cast to be polled.

Let's now take the above audit specification and run it! With the above output in a file called audit.txt, we do this with the command:

./irvaudit -act_ballots Data_NA_Albury.txt_ballots.txt -rep_ballots Data_NA_Albury.txt_ballots.txt -r 0.05 -run audit.txt -s 39284575696785 -eseed 9284896769459 -eprob 0.01

NOTE that you might beed to replace a line of code in model.cpp (line 63) with a version of python you have (if you don't have python3.6. Yes, I know -- this could have been done better. 

You should see a lot of text logging changes in the statistics being maintained by the audit, with the following at the bottom:

TOTAL BALLOTS POLLED = 176 (0.379842%)
SUCCESSFUL AUDITS = 10/10