MEA and temere.epd

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

User avatar
Werner
Posts: 2876
Joined: Wed Mar 08, 2006 10:09 pm
Location: Germany
Full name: Werner Schüle

Re: MEA and temere.epd

Post by Werner »

Hi Ferdy,
is this the same problem ? I tried to use LCZero 0.23.2:
[ 2020-04-25 18:18:12,414 - INFO ] Run engine 58613
[ 2020-04-25 18:18:12,423 - DEBUG ] >> uci
[ 2020-04-25 18:18:12,431 - DEBUG ] << _
[ 2020-04-25 18:18:12,431 - DEBUG ] << | _ | |
[ 2020-04-25 18:18:12,431 - DEBUG ] << |_ |_ |_| v0.23.2+git.c8d9095 built Dec 31 2019
[ 2020-04-25 18:18:12,434 - DEBUG ] << id name Lc0 v0.23.2+git.c8d9095
[ 2020-04-25 18:18:12,434 - DEBUG ] << id author The LCZero Authors.
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name WeightsFile type string default <autodiscover>
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name Backend type combo default blas var blas var random var check var roundrobin var multiplexing var demux
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name BackendOptions type string default
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name Threads type spin default 2 min 1 max 128
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name NNCacheSize type spin default 200000 min 0 max 999999999
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name MinibatchSize type spin default 256 min 1 max 1024
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name MaxPrefetch type spin default 32 min 0 max 1024
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name LogitQ type check default false
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name CPuct type string default 3.000000
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name CPuctBase type string default 19652.000000
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name CPuctFactor type string default 2.000000
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name Temperature type string default 0.000000
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name TempDecayMoves type spin default 0 min 0 max 100
[ 2020-04-25 18:18:12,434 - DEBUG ] << option name TempCutoffMove type spin default 0 min 0 max 1000
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name TempEndgame type string default 0.000000
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name TempValueCutoff type string default 100.000000
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name TempVisitOffset type string default 0.000000
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name DirichletNoise type check default false
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name VerboseMoveStats type check default false
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name FpuStrategy type combo default reduction var reduction var absolute
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name FpuValue type string default 1.200000
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name FpuStrategyAtRoot type combo default same var reduction var absolute var same
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name FpuValueAtRoot type string default 1.000000
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name CacheHistoryLength type spin default 0 min 0 max 7
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name PolicyTemperature type string default 2.200000
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name MaxCollisionEvents type spin default 32 min 1 max 1024
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name MaxCollisionVisits type spin default 9999 min 1 max 1000000
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name OutOfOrderEval type check default true
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name StickyEndgames type check default true
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name SyzygyFastPlay type check default true
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name MultiPV type spin default 1 min 1 max 500
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name PerPVCounters type check default false
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name ScoreType type combo default centipawn var centipawn var centipawn_2018 var win_percentage var Q
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name HistoryFill type combo default fen_only var no var fen_only var always
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name ShortSightedness type string default 0.000000
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name SyzygyPath type string default
[ 2020-04-25 18:18:12,435 - DEBUG ] << option name Ponder type check default true
[ 2020-04-25 18:18:12,436 - DEBUG ] << option name UCI_Chess960 type check default false
[ 2020-04-25 18:18:12,436 - DEBUG ] << option name UCI_ShowWDL type check default false
[ 2020-04-25 18:18:12,436 - DEBUG ] << option name ConfigFile type string default lc0.config
[ 2020-04-25 18:18:12,436 - DEBUG ] << option name KLDGainAverageInterval type spin default 100 min 1 max 10000000
[ 2020-04-25 18:18:12,436 - DEBUG ] << option name MinimumKLDGainPerNode type string default 0.000000
[ 2020-04-25 18:18:12,436 - DEBUG ] << option name SmartPruningFactor type string default 1.330000
[ 2020-04-25 18:18:12,436 - DEBUG ] << option name RamLimitMb type spin default 0 min 0 max 100000000
[ 2020-04-25 18:18:12,436 - DEBUG ] << option name MoveOverheadMs type spin default 200 min 0 max 100000000
[ 2020-04-25 18:18:12,436 - DEBUG ] << option name Slowmover type string default 1.000000
[ 2020-04-25 18:18:12,436 - DEBUG ] << option name ImmediateTimeUse type string default 1.000000
[ 2020-04-25 18:18:12,436 - DEBUG ] << option name LogFile type string default
[ 2020-04-25 18:18:12,436 - DEBUG ] << uciok
[ 2020-04-25 18:18:12,436 - DEBUG ] >> setoption name threads value 1
[ 2020-04-25 18:18:12,436 - DEBUG ] >> setoption name Hash value 64
[ 2020-04-25 18:18:12,436 - INFO ] eoption: ['Slowmover=0']
[ 2020-04-25 18:18:12,436 - DEBUG ] >> setoption name Slowmover value 0
[ 2020-04-25 18:18:12,436 - DEBUG ] >> isready
[ 2020-04-25 18:18:12,436 - DEBUG ] << readyok
[ 2020-04-25 18:18:12,436 - INFO ]

[ 2020-04-25 18:18:12,436 - INFO ] Pos 1
[ 2020-04-25 18:18:12,436 - INFO ] EPD: 1b1r2k1/1R4p1/2n4p/2P1p2P/4N3/6PK/1B6/8 b - - bm Na5; c0 "Rc8=10, Na5=5, Kf8=0, Rf8=0";
[ 2020-04-25 18:18:12,437 - INFO ] id None
[ 2020-04-25 18:18:12,437 - INFO ] FEN: 1b1r2k1/1R4p1/2n4p/2P1p2P/4N3/6PK/1B6/8 b - - 0 1
[ 2020-04-25 18:18:12,437 - INFO ] Solutions: Rc8=10, Na5=5, Kf8=0, Rf8=0
[ 2020-04-25 18:18:12,437 - DEBUG ] >> ucinewgame
[ 2020-04-25 18:18:12,437 - DEBUG ] >> isready
[ 2020-04-25 18:18:12,437 - DEBUG ] << readyok
[ 2020-04-25 18:18:12,437 - DEBUG ] >> position fen 1b1r2k1/1R4p1/2n4p/2P1p2P/4N3/6PK/1B6/8 b - - 0 1
[ 2020-04-25 18:18:12,437 - DEBUG ] >> go movetime 1
[ 2020-04-25 18:18:28,993 - INFO ] 1b1r2k1/1R4p1/2n4p/2P1p2P/4N3/6PK/1B6/8 b - - bm None; ce -32000; acd 0;
[ 2020-04-25 18:18:28,993 - INFO ]

[ 2020-04-25 18:18:28,993 - INFO ] Pos 2
[ 2020-04-25 18:18:28,993 - INFO ] EPD: 1b1r2k1/8/8/1p5p/2pP3r/1P3Pp1/2R1N1P1/3R2K1 b - - bm cxb3; c0 "cxb3=10, Bf4=0, Be5=0, Re8=0";
[ 2020-04-25 18:18:28,993 - INFO ] id None
[ 2020-04-25 18:18:28,993 - INFO ] FEN: 1b1r2k1/8/8/1p5p/2pP3r/1P3Pp1/2R1N1P1/3R2K1 b - - 0 1
[ 2020-04-25 18:18:28,993 - INFO ] Solutions: cxb3=10, Bf4=0, Be5=0, Re8=0
Werner
Top
User avatar
Rebel
Posts: 7025
Joined: Thu Aug 18, 2011 12:04 pm
Full name: Ed Schröder

Re: MEA and temere.epd

Post by Rebel »

No problem here, make sure MEA.EXE is in the same folder as the Lc0 files.

Or in other words, install Lc0 in the TEMERE folder, not in the engines folder.
90% of coding is debugging, the other 10% is writing bugs.
Top
User avatar
pohl4711
Posts: 2459
Joined: Sat Sep 03, 2011 7:25 am
Location: Berlin, Germany
Full name: Stefan Pohl

Re: MEA and temere.epd

Post by pohl4711 »

Rebel wrote: ↑Sat Apr 25, 2020 7:38 pm No problem here, make sure MEA.EXE is in the same folder as the Lc0 files.

Or in other words, install Lc0 in the TEMERE folder, not in the engines folder.
??? lc0 0.24.1 works fine in the engines-folder, where it is in another folder engines\lc0_1. Why should older lc0 or FatFritz not work there??? makes no sense to me. And the logfiles show, that MEA finds and starts FatFritz and older lc0...makes no sense at all.
Top
User avatar
Rebel
Posts: 7025
Joined: Thu Aug 18, 2011 12:04 pm
Full name: Ed Schröder

Re: MEA and temere.epd

Post by Rebel »

pohl4711 wrote: ↑Sat Apr 25, 2020 7:57 pm
Rebel wrote: ↑Sat Apr 25, 2020 7:38 pm No problem here, make sure MEA.EXE is in the same folder as the Lc0 files.

Or in other words, install Lc0 in the TEMERE folder, not in the engines folder.
??? lc0 0.24.1 works fine in the engines-folder, where it is in another folder engines\lc0_1. Why should older lc0 or FatFritz not work there??? makes no sense to me
It might make no sense to you, but have you tried?

Did the job for me, ran into the same problem months ago.
90% of coding is debugging, the other 10% is writing bugs.
Top
User avatar
Werner
Posts: 2876
Joined: Wed Mar 08, 2006 10:09 pm
Location: Germany
Full name: Werner Schüle

Re: MEA and temere.epd

Post by Werner »

Tried here,
thanks Ed - test is running.

stay health
Werner
Werner
Top
Dann Corbit
Posts: 12564
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: MEA and temere.epd

Post by Dann Corbit »

It is astonishing to me that it works so well.
For about 1/3 of the positions, I have a different best move in my database!
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
Top
User avatar
Rebel
Posts: 7025
Joined: Thu Aug 18, 2011 12:04 pm
Full name: Ed Schröder

Re: MEA and temere.epd

Post by Rebel »

Dann Corbit wrote: ↑Sat Apr 25, 2020 9:35 pm It is astonishing to me that it works so well.
For about 1/3 of the positions, I have a different best move in my database!
And it's get better and better after improving the formula, examples:

The Komodo 10 "Dynamism" setting, default setting is 110, tested 7 different setting at 100ms, 45,000 positions, total run time 1 hour, 22 minutes. How long do 7 eng-eng matches of 10,000, 20,000, 30,000 games take?

http://rebel13.nl/mea/K10-Dynamism-100ms.html

-----

Stockfish 10 "Mobility" setting, default setting is 100, tested 7 different setting at 100ms, 250ms and 500ms, 10,000 positions, total run time 21 minutes (100ms), 46 minutes (250ms) and 1:28 (500ms). How long do 7x3 eng-eng matches of 10,000, 20,000, 30,000 games take?

http://rebel13.nl/mea/sf10-ccrl-tb=4-100ms.html
http://rebel13.nl/mea/sf10-ccrl-tb=4-250ms.html
http://rebel13.nl/mea/sf10-ccrl-tb=4-500ms.html

-----

ProDeo "Attractiveness"setting, default setting is 101, tested 7 different setting at 500ms, 4900 positions, total run time 43 minutes. How long do 7 eng-eng matches of 10,000, 20,000, 30,000 games take?

http://rebel13.nl/mea/Attractiveness-500ms.html
90% of coding is debugging, the other 10% is writing bugs.
Top
Dann Corbit
Posts: 12564
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: MEA and temere.epd

Post by Dann Corbit »

It must be one of those "laws of big numbers" things that makes it work so well.
If it can be used to make engines play better, then it is revolutionary.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
Top
User avatar
pohl4711
Posts: 2459
Joined: Sat Sep 03, 2011 7:25 am
Location: Berlin, Germany
Full name: Stefan Pohl

Re: MEA and temere.epd

Post by pohl4711 »

Rebel wrote: ↑Sat Apr 25, 2020 8:00 pm
pohl4711 wrote: ↑Sat Apr 25, 2020 7:57 pm
Rebel wrote: ↑Sat Apr 25, 2020 7:38 pm No problem here, make sure MEA.EXE is in the same folder as the Lc0 files.

Or in other words, install Lc0 in the TEMERE folder, not in the engines folder.
??? lc0 0.24.1 works fine in the engines-folder, where it is in another folder engines\lc0_1. Why should older lc0 or FatFritz not work there??? makes no sense to me
It might make no sense to you, but have you tried?

Did the job for me, ran into the same problem months ago.
I do not doubt any of your words. What I ment was, that I have absolutely no idea, what kind of bug in MEA could lead to such a strange behaviour: lc0 0.23.2 and FatFritz do not run in the engines folder (even though the logfile show, that the engine is started by MEA) and lc0 0.24.1 works fine in the engines folder ?!
Top
User avatar
pohl4711
Posts: 2459
Joined: Sat Sep 03, 2011 7:25 am
Location: Berlin, Germany
Full name: Stefan Pohl

Re: MEA and temere.epd

Post by pohl4711 »

Dann Corbit wrote: ↑Sun Apr 26, 2020 12:16 am It must be one of those "laws of big numbers" things that makes it work so well.
If it can be used to make engines play better, then it is revolutionary.
It is! And it is perfect for fast tests (only 1 calculated node!) of neural-nets...
Look on my website:
https://www.sp-cc.de/nn-mea-testing.htm

Example:

Code: Select all

Engine                           :  Top1  Top1Rate  Score  ScoreRate
lc0 0.24.1 LS 14.3 (20x256)      : 17929    0.515  257442    0.739
lc0 0.24.1 LS 14.2 (20x256)      : 17899    0.514  257012    0.738
lc0 0.24.1 LS 14.1 (20x256)      : 17769    0.510  255712    0.734
lc0 0.24.1 LS 14 (20x256)        : 17749    0.509  255361    0.733
Not only the ranking of all LS 14 nets is correct. Aditionally, the wider gap between LS 14.1 and LS 14.2 is correct. Awesome!
Top

Return to “Computer Chess Club: General Topics”