After all my NNUE experiments, it's time to release a new version of my little engine Orion !
The main changes are :
- the use of a "NNUE-like" neural network as evaluation function ;
- the release of both my trainer (Python) and inference (C langage) source code as a "library".
More information on my website (see signature)
Download the Orion chess engine --- Train your NNUE with the Cerebrum library --- Contribute to the Nostradamus experiment !
To late for my bigger and still running FCP Qualify Tourney-2021.
But v0.7 is playing.
Good luck with your Orion and have thanks for your new version!
Looking on weakend often on your site!
I hope available for my tourney, but it comes to late for me.
Thanks a lot to both of you for your kind words. The training of a NN is a real challenge !
@Gabor : this time, I promise, there won't be a "@CCRL team : oh no, no, no !" topic, as this is an "official" version of Orion
@Frank : yes I saw that is was too late, but I wanted to wait for the end of my tests before releasing. I must admit that I'm not comfortable with the fact that the "experimental" version of Orion 0.7 will play, but for sure you're free. Any chance to replace it by the "official" 0.8 version ?
Download the Orion chess engine --- Train your NNUE with the Cerebrum library --- Contribute to the Nostradamus experiment !
I can make an execption because I am the master of disaster.
After my late breakfast today I will stopped the tourney and changed that if you like to see your new version in my tourney. Games from Orion 0.7 will be replayed.
But please ...
Do not tell that anyone from the others!
changed, with News-08 on detail page and a new entry for my new configuration site.
5 games to replay.
I think this is for my tourney the clearly better solution because at first I added
the v0.7 __beta__ version. And to see that the network file comes not from Stockfish
is clearly more interesting for all the others.
Thank you again and good luck for Orion 0.8 ...
Go Orion go ...
I see that you removed syzygy support. Did you do anything special with regards to training in late endgame positions? I have played with Stockfish's NNUE nets and code in my own weak engine, and found that it shuffled in some late endgame positions, though that might be due in part to lack of search depth.
tmokonen wrote: ↑Tue Dec 01, 2020 8:24 pm
I see that you removed syzygy support. Did you do anything special with regards to training in late endgame positions? I have played with Stockfish's NNUE nets and code in my own weak engine, and found that it shuffled in some late endgame positions, though that might be due in part to lack of search depth.
No, nothing special. But there were endgame positions in the set used for training. The idea was to see how a single and "small" NN would perform alone, without any endgame knowledge access (except 3-pieces endgames for which I use an own-made KPK bitbase).
Download the Orion chess engine --- Train your NNUE with the Cerebrum library --- Contribute to the Nostradamus experiment !
Frank Quisinsky wrote: ↑Tue Dec 01, 2020 11:19 am
And to see that the network file comes not from Stockfish
is clearly more interesting for all the others.