Final Release of Ethereal, V12.75

[chrisw]

first we lost Gull and Fizbo... after that, Andscacs... it seems Xiphos is not coming back either... and now Ethereal...

people get scared too easily from this neural monsters like leela and SF NNUE... This shows that more effort needs to be put in eval functions, and not only in perfecting search...

Don't you think it was SF rather than SF+NNUE that most got scared off? How can anyone compete with the whole SF team plus fishtest and its huge computational resources?

Kuhn. History of Scientific Revolutions.

[AndrewGrant]

https://github.com/AndyGrant/Ethereal/tree/nnue
Pushed NNUE support to a branch. You can load in a file. Stockfish weights are preferred.
If anyone tests this for a rating list, it should be called "Ethereal 12.75 SF-NNUE".
That branch actually works, unlike the Etherlito crap which broke due to memory alignments on POSIX.

[jpqy]

Your last source show this after compiling

uci
id name Ethereal 12.69 (POPCNT)
id author Andrew Grant, Alayan & Laldon
option name Hash type spin default 16 min 2 max 131072
option name Threads type spin default 1 min 1 max 2048
option name MultiPV type spin default 1 min 1 max 256
option name EvalFile type string default <empty>
option name MoveOverhead type spin default 100 min 0 max 10000
option name SyzygyPath type string default <empty>
option name SyzygyProbeDepth type spin default 0 min 0 max 127
option name Ponder type check default false
option name AnalysisMode type check default false
option name UCI_Chess960 type check default false
uciok

[Rebel]

Your last source show this after compiling

uci
id name Ethereal 12.69 (POPCNT)
id author Andrew Grant, Alayan & Laldon
option name Hash type spin default 16 min 2 max 131072
option name Threads type spin default 1 min 1 max 2048
option name MultiPV type spin default 1 min 1 max 256
option name EvalFile type string default <empty>
option name MoveOverhead type spin default 100 min 0 max 10000
option name SyzygyPath type string default <empty>
option name SyzygyProbeDepth type spin default 0 min 0 max 127
option name Ponder type check default false
option name AnalysisMode type check default false
option name UCI_Chess960 type check default false
uciok

And I was afraid Andy would call it EvilFile, I was wrong, fortunately

[frankp]

Not sure what I did wrong. Increasing threads slows down the search - time to depth and nps. (Ryzen 2700X stock.).

./Ethereal
go depth 20
info depth 20 seldepth 31 multipv 1 score cp 41 time 12172 nodes 3253344 nps 267000 tbhits 0 hashfull 487 pv d2d4 d7d5 c2c4 e7e6 g1f3 g8f6 b1c3 a7a6 c4d5 e6d5 c1f4 c7c5 f4g5 b8c6 g5f6 g7f6 d4c5 d5d4 c3e4 f6f5 e4d6 f8d6 c5d6 d8b6 a1b1 c8e6

//---------------------------------------------------------------------------
setoption name Threads value 16
info string set Threads to 16
info depth 18 seldepth 29 multipv 1 score cp 28 upperbound time 94268 nodes 16092460 nps 170000 tbhits 0 hashfull 961 pv
info depth 18 currmove d2d4 currmovenumber 1
stop

//----------------------------------------------------------------------------
setoption name Threads value 2
info string set Threads to 2
go depth 20
info depth 20 seldepth 28 multipv 1 score cp 30 time 53169 nodes 6323333 nps 118000 tbhits 0 hashfull 752 pv d2d4 d7d5 c2c4 e7e6 g1f3 g8f6 b1c3 f8e7 c1g5 h7h6 g5f6 e7f6 e2e3 e8g8 f1e2 c7c5 c4d5 e6d5 d4c5 b8c6 e1g1 c8e6 d1b3

//---------------------------------------------------------------------------------
setoption name Threads value 8
info string set Threads to 8
go depth 20
info depth 18 seldepth 27 multipv 1 score cp 33 time 73747 nodes 8660783 nps 117000 tbhits 0 hashfull 812 pv d2d4 d7d5 c2c4 e7e6 b1c3 a7a6 c4d5 e6d5 g1f3 h7h6 e2e3 g8f6 f1d3 f8d6 e1g1 e8g8 d1b3 c7c6 c1d2
info depth 19 currmove d2d4 currmovenumber 1
stop
bestmove d2d4 ponder d7d5

[AndrewGrant]

Not sure what I did wrong. Increasing threads slows down the search - time to depth and nps. (Ryzen 2700X stock.).

No clue. Make sure you don't have other processes running that are taking the entire CPU? See if you get the same results with non NNUE Ethereal?

Kicked up the version number that was pointed out as being wrong.

[frankp]

htop showed just Ethereal running on chosen number of threads - taking >99.x cpu resource.
Umm.....

OK recompiled and seems fine.
Think I may have (previously) run 'profile' compile. Ouch.

[syzygy]

SF doesn't have a closed team of developers. Dozens of people have contributed to it, and the NNUE authors are among them.

Perhaps we have a different understanding of what that means.

Apparently we have, even though what I wrote is not very complex.

I guess in your view 99% of Stockfish is not an achievement of Stockfish. That's fine, I just don't think it is a very useful view. Stockfish is not the effort of a closed team of contributors.

Btw, if I am not mistaken, the original Shogi NNUE engine was based on Stockfish.

[chysiddh14]

I m vry vry sad by this final release .
i think many people will join ethereal n make it stronger no. 1 chess engine in the world .
i m hoping for this success.

plz come back andrew,alayan make it more powerful .
even ur hobby is making ur name n ethereal remarkable.

[jpqy]

Thanks ,looks right now!

uci
id name Ethereal 12.75 NNUE (POPCNT)
id author Andrew Grant, Alayan & Laldon
option name Hash type spin default 16 min 2 max 131072
option name Threads type spin default 1 min 1 max 2048
option name MultiPV type spin default 1 min 1 max 256
option name EvalFile type string default <empty>
option name MoveOverhead type spin default 100 min 0 max 10000
option name SyzygyPath type string default <empty>
option name SyzygyProbeDepth type spin default 0 min 0 max 127
option name Ponder type check default false
option name AnalysisMode type check default false
option name UCI_Chess960 type check default false
uciok

Your normal Ethereal 12.75 also doing great in first tests!