fishtest
fishtest copied to clipboard
glinscott
Stockfish testing
As the redistribution of packages inside our repo appears necessary for niche technical reasons, include all the root licenses of the packages we distribute with the code. closes: https://github.com/official-stockfish/fishtest/issues/1873
The previous message is confusing and vague enough to convey the reversed meaning.
LLR calculation is not in native normalized Elo (first normalized bounds are converted to logistic bounds, then log-likelihoods are computed as posteriors to uniform conditional to expected value matching logistic...
Python packages are not fishtest code, and should not be a part of the repo for couple of reasons: 1- Copywrite concerns 2- Bloated repo (upload, download) 3- Creates confusion
Every time the local worker is started for a new batch of games, it must pull the sources of both repos and compile. However, often the same repos are pulled...
Scale itp calculation based on total throughput instead of active runs count. This distributes the number of cores more fairly between users, and allows for an effective way to transfer/split...
closes https://github.com/official-stockfish/fishtest/pull/1827
fishset stuck in loop wont create any file access denied by opening worker.py   
Separate issue for #1781 - Problem: Windows cannot handle large command lines See: https://tests.stockfishchess.org/actions?action=&user=mgrabiak&text=The+filename+or+extension+is+too+long - Windows worker shouldnt be assigned to these tests. - Windows max length seems to be...
I wonder if there is any appetite for auto blocking workers under some conditions. For example after two (three?) consecutive crash_or_time events.
