There isn’t any global/overall summary data about the number of decoys, but you can usually get the information out with grep & wc. You just need to pick a line which occurs once per decoy, and count the number of lines. I usually do “grep SCORE silent.out | wc -l”, but that has one more than the number of structures, due to the header score line. You may prefer “grep ANNOTATED_SEQUENCE silent.out | wc -l”. Note that that doesn’t count how may structures are valid – if you application stopped in the middle there might be a corrupted structure which has a SCORE/ANNOTATED_SEQUENCE line but isn’t readable. You’d have to use something like the score_jd2 application with the “-silent_read_through_errors” flag to count how many structures are valid in the silent file. (It should print the number in the log file, as well as having one line per structure in the output score file.) This will be slower than the grep/wc route, though, especially with a large number of decoys.
By the way, Rosetta is generally set up to be robust to accidentally halted applications. You normally should be able to restart with the exact same commandline and have Rosetta pick up where it left off.