Explanation of the contents of the results file "out.txt"?

[mathisdon]

Hi, I am able to run abcrown.py in Linux on the tutorial_examples, but I can't find an explanation of what I see in the file "out.txt". I unpickled the file and can read it, but I have questions.

1. What determines how many lines there will be? My run of onnx_with_built-in_dataset_linf_bound.yaml is up to 1681 lines so far.

2. What is the results format?
   'results': [['safe-incomplete', 3.5424506664276123],
   ['unsafe-pgd', 0.51487135887146],
   ['safe-incomplete', 0.6974430084228516],
   ['unsafe-pgd', 0.007823705673217773],
   ['safe-incomplete', 0.6528849601745605],
   ... (>1600 more lines)

3. {'bab_ret': [[14,
   [[3.4012045860290527]],
   0,
   0.16716694831848145,
   array([[0]], dtype=int64)],
   [14, 1.0000000116860974e-07, 4, 3.8302958011627197, [0]],
   [14,
   [[3.4012045860290527]],
   0,

4. 'summary': defaultdict(<class 'list'>,
   {'safe': [14,
   112,
   194,
   341,
   413,
   787,
   901,
   1332,
   1427,
   1568],
   'safe-incomplete': [0,
   2,
   4,
   5,
   6,
   9,
   ...

5. And a related question: How do I tell abcrown.py which algorithm to run? Can I tell it to only compute bounds? Or to only find couterexamples?