Discussion of chess software programming and technical issues.
Moderator: Ras
Daniel Shawul
Posts: 4186Joined: Tue Mar 14, 2006 11:34 amLocation: Ethiopia
Post by Daniel Shawul Sat Oct 17, 2020 5:27 pm
maksimKorzh wrote: ↑ Sat Oct 17, 2020 10:42 am
Hi guys, thanks to amazing library to probe stockfish NNUE by Daniel Shawul even a noob like me can now get NNUE score for each position, but it turns out that using that score directly leads to a VERY STRANGE behavior of my engine = it makes strange moves from time to time and sacrifices the queen for nothing in every game. I feel a bit confused at this point.
My questions are:
1. Do I need to evaluate material if I want to use pure NNUE eval?
2. Do I need to scale NNUE eval?
3. Do I need to tune material weights? (I heard SF uses around 200 for paen value while my engine does around 100)
Actually what should I do with NNUE score to make my engine play normal chess?
Please help.
FYI I have now added support for the incremental evaluation of NNUE in the library.
This gives a speed bump of 14 to 18%, so maybe worth it for those in need of the highest performance.
The engine has to keep track of dirtyPiece and Accumulator using the same data structure as what the library use.
Make sure accumulator is cache aligned i.e. 64 bytes.
It maybe possible to move some of the code for incremental evaluation from the engine to the library in the future.
For an example implementation of "NNUE incremental" in an engine, look here
https://github.com/dshawul/Scorpio/comm ... 7b7322bf84
If you want "NNUE without incremental evalution", the implementation is much easier
https://github.com/dshawul/Scorpio/comm ... 94cd4b2c5c
I got more that 400 elos out of my implementation with a shared library.
maksimKorzh
Posts: 775Joined: Sat Sep 08, 2018 5:37 pmLocation: UkraineFull name: Maksim Korzh
Post by maksimKorzh Sat Oct 17, 2020 6:49 pm
hgm wrote: ↑ Sat Oct 17, 2020 2:35 pm
Let me get this straight: NNUE is just one term of the Stockfish evaluation, and there are other terms?
Is material the only other term (and is it tapered from opening to end-game)?
Yes, material is the ONLY other term and YES it's tapered from opening to end-game.
maksimKorzh
Posts: 775Joined: Sat Sep 08, 2018 5:37 pmLocation: UkraineFull name: Maksim Korzh
Post by maksimKorzh Sat Oct 17, 2020 6:53 pm
Daniel Shawul wrote: ↑ Sat Oct 17, 2020 5:27 pm maksimKorzh wrote: ↑ Sat Oct 17, 2020 10:42 am
Hi guys, thanks to amazing library to probe stockfish NNUE by Daniel Shawul even a noob like me can now get NNUE score for each position, but it turns out that using that score directly leads to a VERY STRANGE behavior of my engine = it makes strange moves from time to time and sacrifices the queen for nothing in every game. I feel a bit confused at this point.
My questions are:
1. Do I need to evaluate material if I want to use pure NNUE eval?
2. Do I need to scale NNUE eval?
3. Do I need to tune material weights? (I heard SF uses around 200 for paen value while my engine does around 100)
Actually what should I do with NNUE score to make my engine play normal chess?
Please help.
FYI I have now added support for the incremental evaluation of NNUE in the library.
This gives a speed bump of 14 to 18%, so maybe worth it for those in need of the highest performance.
The engine has to keep track of dirtyPiece and Accumulator using the same data structure as what the library use.
Make sure accumulator is cache aligned i.e. 64 bytes.
It maybe possible to move some of the code for incremental evaluation from the engine to the library in the future.
For an example implementation of "NNUE incremental" in an engine, look here
https://github.com/dshawul/Scorpio/comm ... 7b7322bf84
If you want "NNUE without incremental evalution", the implementation is much easier
https://github.com/dshawul/Scorpio/comm ... 94cd4b2c5c
I got more that 400 elos out of my implementation with a shared library.
Thanks Daniel!
I've annoyed you with couple more issues on github...
Can you please clarify one more question - I still can't figure it out from CFish's source:
Should I use pawn material weight relatively equal to 100 or 200
the reason why I'm asking is here:
[url]view-source:
https://hxim.github.io/Stockfish-Evaluation-Guide/ [/url]
see line 1153
It says
and this confuses me... may be it's specific to stockgish evaluation guide GUI...
hgm
Posts: 28468Joined: Fri Mar 10, 2006 10:06 amLocation: AmsterdamFull name: H G Muller
Post by hgm Sat Oct 17, 2020 6:56 pm
OK, thanks. I ask because I am interested in trying out this NNUE technique in a new engine, for Janggi. I know absolutely nothing about that game, but I do have a PGN file with 25000 games in it. And perhaps this is enough to train it, and get somewhat reasonable play.
maksimKorzh
Posts: 775Joined: Sat Sep 08, 2018 5:37 pmLocation: UkraineFull name: Maksim Korzh
Post by maksimKorzh Sat Oct 17, 2020 7:15 pm
hgm wrote: ↑ Sat Oct 17, 2020 6:56 pm
OK, thanks. I ask because I am interested in trying out this NNUE technique in a new engine, for Janggi. I know absolutely nothing about that game, but I do have a PGN file with 25000 games in it. And perhaps this is enough to train it, and get somewhat reasonable play.
It could also be interesting to try it for Paco Shako.
I am always getting excited when you're trying to implement engines for games other but chess. I'm just wondering what motivates you?
I mean usually people developing one engine for years and it's difficult to master chess programming at a decent level...
But you're trying many other games! This is incredible, so what motivates you? Thanks in advance.
maksimKorzh
Posts: 775Joined: Sat Sep 08, 2018 5:37 pmLocation: UkraineFull name: Maksim Korzh
Post by maksimKorzh Sat Oct 17, 2020 8:41 pm
After a consultation with Daniel Shawul and fixing the bugs I can now say for sure:
Daniel Shawul
Posts: 4186Joined: Tue Mar 14, 2006 11:34 amLocation: Ethiopia
Post by Daniel Shawul Sat Oct 17, 2020 9:21 pm
Some may want to do scaling to get better compatibility with search prunings such as
hgm
Posts: 28468Joined: Fri Mar 10, 2006 10:06 amLocation: AmsterdamFull name: H G Muller
Post by hgm Sat Oct 17, 2020 9:23 pm
maksimKorzh wrote: ↑ Sat Oct 17, 2020 7:15 pm I am always getting excited when you're trying to implement engines for games other but chess. I'm just wondering what motivates you?
I mean usually people developing one engine for years and it's difficult to master chess programming at a decent level...
But you're trying many other games! This is incredible, so what motivates you? Thanks in advance.
I am driven mainly by curiosity, I like doing creative work, and it is gratifying to create things that others are eager to use. That sort of disqualifies orthodox Chess, for which there are already thousands of engines. Except for a handful of testers no one is interested in having yet another one. So I try to focus on games where nothing, or nothing decent exists yet. Or where it is truly unknown what is a good way to do it.
I find Paco Shako a very intriguing game in this respect; no one really knows how to play that. I am very curious what would be a good strategy for it. This NNUE method seems ideally suited for it, and material does not seem to be an important term in that game (it is impossible to lose any...), and King Safety is everything. And as I happen to have a lot of Janggi games handy, this seems a good opportunity to get some hands-on experience with this interesting NNUE technique. For Janggi there seems to hardly exist anything, and millions of potential users.
maksimKorzh
Posts: 775Joined: Sat Sep 08, 2018 5:37 pmLocation: UkraineFull name: Maksim Korzh
Post by maksimKorzh Sat Oct 17, 2020 11:54 pm
hgm wrote: ↑ Sat Oct 17, 2020 9:23 pm maksimKorzh wrote: ↑ Sat Oct 17, 2020 7:15 pm I am always getting excited when you're trying to implement engines for games other but chess. I'm just wondering what motivates you?
I mean usually people developing one engine for years and it's difficult to master chess programming at a decent level...
But you're trying many other games! This is incredible, so what motivates you? Thanks in advance.
I am driven mainly by curiosity, I like doing creative work, and it is gratifying to create things that others are eager to use. That sort of disqualifies orthodox Chess, for which there are already thousands of engines. Except for a handful of testers no one is interested in having yet another one. So I try to focus on games where nothing, or nothing decent exists yet. Or where it is truly unknown what is a good way to do it.
I find Paco Shako a very intriguing game in this respect; no one really knows how to play that. I am very curious what would be a good strategy for it. This NNUE method seems ideally suited for it, and material does not seem to be an important term in that game (it is impossible to lose any...), and King Safety is everything. And as I happen to have a lot of Janggi games handy, this seems a good opportunity to get some hands-on experience with this interesting NNUE technique. For Janggi there seems to hardly exist anything, and millions of potential users.
Thanks for explanations. Very interesting! It would be even more interesting if you were making microMax like tutorials - I've learned so much from that tutorial! Your answers here are for good programmers (not for me) but your tutorials are great!
maksimKorzh
Posts: 775Joined: Sat Sep 08, 2018 5:37 pmLocation: UkraineFull name: Maksim Korzh
Post by maksimKorzh Tue Oct 20, 2020 9:20 pm
I'm facing some very weird behavior of Stockfish NNUE in my engine:
