IbaiBuR wrote: Wed Jun 21, 2023 3:54 pm
Just a simple question. Why do you use the Self-Q-Learning option enabled with BrainLearn to play a match against other engines? This, actually, makes no sense. Just to let you know, the QLearning is a Reinforcement Learning based algorithm which is mainly optimized for
selfplay. So, it is normal that BrainLearn's performance is weird since you are using it incorrectly...
The reason,
BrainLearn v24 (with SQL enabled) is generated Experience file
Otherwise, (when BrainLearn engine with default settings),
No any Experience file (during or after game-play) recorded, at least on my machine!
Maybe the Experience file is located in other folder? I mean when default settings...
I do not think so that to be located in other folder....but maybe I am missing something?
Plus, do you know about what is the right setting for Brainlearn (about Learning On)?
Thanks in advance
Eman is actually not designed to be used with parallel matches, thats why it crashed basically, probably even the experience file got corrupted...
Really ? anyhow, I do not believe that that Eman exp file is got corrupted...
If it was corrupted..then I hardly doubt that Eman to win the recent tour + Duel..
And I hope also that, sure in the next releases,
Eman will be optimized for parallel matches too...
Because so far
All other rest Top SF based engines are stable with concurrent: 64
Because the stability is the most important part...at least for my side!