Hey what's up guys, Code Monkey King's here.
I've live coded the tutorial on how to embedd latest Stockfish NNUE to legendary VICE chess engine by Bluefever Software:
https://www.youtube.com/watch?v=QJJXhv5wDFk
SOURCE CODE:
https://github.com/maksimKorzh/vice-nnue
Linux binary exacutable:
https://github.com/maksimKorzh/vice-nnu ... ource/vice
Windows binary executable:
https://github.com/maksimKorzh/vice-nnu ... e/vice.exe
NN file:
https://github.com/maksimKorzh/vice-nnu ... 53044.nnue
NNUE probe library by Daniel Shawul:
https://github.com/dshawul/nnue-probe
VICE chess engine on NNUE steroids!
Moderators: hgm, Dann Corbit, Harvey Williamson
Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
- maksimKorzh
- Posts: 627
- Joined: Sat Sep 08, 2018 3:37 pm
- Location: Ukraine
- Full name: Maksim Korzh
- Contact:
VICE chess engine on NNUE steroids!
Wukong Xiangqi (Chinese chess engine + apps to embed into 3rd party websites):
https://github.com/maksimKorzh/wukong-xiangqi
Chess programming YouTube channel:
https://www.youtube.com/channel/UCB9-pr ... KKqDgXhsMQ
https://github.com/maksimKorzh/wukong-xiangqi
Chess programming YouTube channel:
https://www.youtube.com/channel/UCB9-pr ... KKqDgXhsMQ
-
- Posts: 568
- Joined: Mon Jul 20, 2015 3:06 pm
- Contact:
Re: VICE chess engine on NNUE steroids!
Thank you for the NNUE example. The code compiles fine. However, the "eval" results look very strange. For example:
[d]3r1k2/4npp1/1ppr3p/p6P/P2PPPP1/1NR5/5K2/2R5 w - -
produces a score of 433. The elo is guessed less than 2500 so it is hardly steroids.
[d]3r1k2/4npp1/1ppr3p/p6P/P2PPPP1/1NR5/5K2/2R5 w - -
produces a score of 433. The elo is guessed less than 2500 so it is hardly steroids.
-
- Posts: 1173
- Joined: Sun Mar 12, 2006 5:46 pm
- Location: Kelowna
- Full name: Tony Mokonen
- Contact:
Re: VICE chess engine on NNUE steroids!
To be fair, the original Vice is around 2050 Elo, so even a 2400 rating for Vice NNUE would be a massive improvement.D Sceviour wrote: ↑Tue Oct 20, 2020 6:11 pmTThe elo is guessed less than 2500 so it is hardly steroids.
- maksimKorzh
- Posts: 627
- Joined: Sat Sep 08, 2018 3:37 pm
- Location: Ukraine
- Full name: Maksim Korzh
- Contact:
Re: VICE chess engine on NNUE steroids!
I've sligtnly updated sources:
1. Added optimization flags in makefile (it now hits almost the same depth as classical version!)
2. Added 50 move rule counter penalty in the evaluation (as it's done in CFish)
Regarding penalty - the idea is simple: instead of return nnue_score we return nnue_eval * (100 - pos->fiftyCount) / 100
In my engine BBC without this penalty it fails to mate simple endgames, but with this feature it crushes VICE 1.0 9,5 : 0,5
Here are a couple of games:
PGN: https://github.com/maksimKorzh/bbc/tree ... s_vice_1.0
1. Added optimization flags in makefile (it now hits almost the same depth as classical version!)
2. Added 50 move rule counter penalty in the evaluation (as it's done in CFish)
Regarding penalty - the idea is simple: instead of return nnue_score we return nnue_eval * (100 - pos->fiftyCount) / 100
In my engine BBC without this penalty it fails to mate simple endgames, but with this feature it crushes VICE 1.0 9,5 : 0,5
Here are a couple of games:
PGN: https://github.com/maksimKorzh/bbc/tree ... s_vice_1.0
Wukong Xiangqi (Chinese chess engine + apps to embed into 3rd party websites):
https://github.com/maksimKorzh/wukong-xiangqi
Chess programming YouTube channel:
https://www.youtube.com/channel/UCB9-pr ... KKqDgXhsMQ
https://github.com/maksimKorzh/wukong-xiangqi
Chess programming YouTube channel:
https://www.youtube.com/channel/UCB9-pr ... KKqDgXhsMQ