Run
New experiment
Pick what to run, then start it here or copy the command. The UI, transitive.py and benchmark.py all run a campaign the same way: one process per run, a time limit, and a check of every result. The results appear under Campaigns.
Which systems?
Leave everything unselected to run every discovered system.
Which topologies?
Leave everything unselected to run every topology.
Settings
Where the results go, recursion modes, graph sizes, repetitions and the time limit.
results/
An existing campaign is resumed: configurations it already has are not run again.
Use letters, digits, dot, dash and underscore.
A run over the limit is killed; larger sizes of that configuration are skipped. The paper uses 600.
Query mode
Compute the whole closure.
The paper averages 5 runs.
Only used when Soufflé is selected.
Review and start
This public copy cannot run benchmarks: copy a command and run it on your machine.
Runs to execute
0
Same run from the command line
With the size range of this form:
$
With the sizes listed one by one (this is how scripts/run_all.sh calls it):
$
Both commands run the same engine as the Start button. Results land in results/<campaign>/<system>/runs.jsonl.
Copy a command above and run it on your machine.