CSS-Shop Alt Div P Results
```
# PLAYER : RATING ERROR POINTS PLAYED (%) CFS(%) W D L D(%) Wscore Bscore Pairs Speed
1 Stockfish 17 : 137 61 39.5 56 70.5 94 25 29 2 51.8 23 16.5 23-0 4.1Mnps
2 Lc0-Bt4-6147500-it332 : 69 56 34.0 56 60.7 81 21 26 9 46.4 19.5 14.5 14-2 2.0Knps
3 Obsidian 14 : 34 52 31.0 56 55.4 83 16 30 10 53.6 17.5 13.5 11-5 6.2Mnps
4 Ceres C1-512-25 : -6 54 27.5 56 49.1 50 12 31 13 55.4 15 12.5 8-9 4.1Knps
5 berserk 13 : -6 55 27.5 56 49.1 55 13 29 14 51.8 16 11.5 8-9 8.2Mnps
6 dragon 3.3 : -11 56 27.0 56 48.2 98 10 34 12 60.7 15 12 8-10 5.8Mnps
7 Ethereal 14.25 : -99 58 19.5 56 34.8 67 6 27 23 48.2 12.5 7 4-21 8.9Mnps
8 Seer 2.80 : -118 55 18.0 56 32.1 --- 3 30 23 53.6 11.5 6.5 0-20 5.9Mnps
White advantage = 0.00 +/- 0.00
Draw rate (equal opponents) = 59.54 % +/- 3.71
```
Auch ich kann bestätigen, dass Rems M-091024 zu den top Engines gehört.
In der Analyse weit besser als Stockfish 17.
Gruss
Kurt
Ceres c1-512-25 is very close / just behind Berserk 13 in strength.
This result bears out the same close matchup from the TCEC Cup
Now just waiting on the release of the new Ceres net - to test for further improvement
```
# PLAYER : RATING ERROR POINTS PLAYED (%) CFS(%) W D L D(%) Wscore Bscore Pairs Speed
1 berserk 13 : 24 12 434.5 800 54.3 63 211 447 142 55.9 292 142.5 132-63 144.6Mnps
2 Ceres C1-512-25 : 19 27 98.5 200 49.3 76 43 111 46 55.5 68 30.5 25-28 44.0Knps
3 Ceres C1-512-25 Tuned : 3 28 94.0 200 47.0 76 41 106 53 53.0 68 30.5 25-28 44.0Knps
4 Ceres-768x15_NLA : -14 30 89.0 200 44.5 77 32 114 54 57.0 62.5 26.5 12-34 47.3Knps
5 Ceres-768x15_NLA Tuned : -32 28 84.0 200 42.0 --- 26 116 58 58.0 62.5 26.5 12-34 47.3Knps
White advantage = 0.00 +/- 0.00
Draw rate (equal opponents) = 56.55 % +/- 1.65
```
Ceres 1.00 C1-640-25
Bisher gelöst: 72 von 114 ; 11:08m
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
-------------------------------------------------------------------------------------
0 | 0 - - - - 0 - - 2 0 7 - 0 0 0 1 0 8 10 0
20 | 0 2 - - 0 0 0 - - - - - - 2 0 0 0 - - 3
40 | 0 0 - 0 - - 11 0 - - - 5 0 0 0 0 0 - - 6
60 | - - 0 0 0 0 - 8 - 4 - - 2 0 - - 6 1 2 -
80 | 0 0 0 0 - 0 0 - 1 0 - 0 - 9 2 1 3 3 - 0
100 | - 0 0 0 - - 0 0 0 0 1 2 0 5
K/s: 2.499
TotTime: 14:40m SolTime: 11:08m EPD : 2112.epd
Time : 1000 ms
Max Total Time Hash
Engine Score Found Pos ELO Score Rate ms Mb Cpu
1 Stockfish241113 110920 1836 2112 4018 124155 89.3% 1000 32 4
2 Stockfish241012 109684 1802 2112 3973 124155 88.3% 1000 32 4
3 Lc0v0.32-6147500 108776 1790 2112 3942 124155 87.6% 1000 100 2
4 Lc0v0.31-6147500 108706 1786 2112 3942 124155 87.6% 1000 100 2
5 Ceres1 C1-640-25 104801 1711 2112 3798 124155 84.4% 1000 32 1
6 Obsidian14 103438 1710 2112 3748 124155 83.3% 1000 32 4
7 Dragon3.3 103313 1703 2112 3744 124155 83.2% 1000 32 4
8 Berserk13 96914 1605 2112 3514 124155 78.1% 1000 32 4
Created with MEA
by
Ferdinand
Mosca
) sein gelassen, was da alles dann erst in der Konsole an weiteren Anforderungen aufgelistet wird (über das Laden eines WeightsFiles hinaus, das ging noch einfach, aber auch nicht automatisch, man fragt sich auch da, warum muss man das händisch machen), hat mich die Sache gleich wieder abdrehen lassen. Wenn ich einen von gut 20 verschiedenen im Netz dauernd kursierenden und laufend aktualisierten SF- Branches neu installiere, muss ich das alles nicht machen, und der Unterschied zu SF dev. ist wahrscheinlich größer als der zwischen Ceres und Lc0, just my two cents about the too sour grapes
) keine Konflikte mit Apps gemacht, die ich schon probiert habe. Danach lässt sich Ceres in der Konsole erstmal starten, beim go infinite- Versuch meckert's zunächst, dass es kein WeightsFile findet, gibt aber auch gleich einen Demo- Befehl dazu an, auf den hinauf lädt's dann was, sagt ok und beim nächsten go infinite kommen ein paar Bildschirme voller error- Meldungen, was es noch alles gern hätte, abgeschlossen mit der Meldung
uciok
setoption name WeightsFile value C1-256-10
uci info Network evaluation configured to use: <NNEvaluatorDef Network=Ceres:C1-256-10 Device=GPU:0 >
isready
Loaded network weights: C1-256-10
Starting ONNX runtime against C1-256-10.onnx from C1-256-10.onnx with GPU 0 using (CUDA FP16)
0,384 sec TC: ONNX InferenceSession create on model of size 73.151.416 bytes mdiff: 0,00 memTot: 0,00 CPU: 0,48 secs, GC 0 0 0 WS: 0,46 WSPeak: 0,54
Starting ONNX runtime against C1-256-10.onnx from C1-256-10.onnx with GPU 0 using (CUDA FP16)
0,387 sec TC: ONNX InferenceSession create on model of size 73.151.416 bytes mdiff: 0,00 memTot: 0,00 CPU: 0,56 secs, GC 1 1 1 WS: 0,63 WSPeak: 0,69
readyok
go nodes 1000
info depth 1 seldepth 1 time 147 nodes 1 score cp 5 tbhits 0 nps 7 pv d2d4 string M= NaN
info depth 3 seldepth 5 time 654 nodes 66 score cp 6 tbhits 0 nps 101 pv d2d4 d7d5 c2c4 c7c6 string M= NaN
info depth 4 seldepth 7 time 1163 nodes 204 score cp 7 tbhits 0 nps 175 pv d2d4 d7d5 c2c4 c7c6 c4d5 c6d5 string M= NaN
info depth 4 seldepth 7 time 1706 nodes 386 score cp 6 tbhits 0 nps 226 pv d2d4 d7d5 c2c4 c7c6 c4d5 c6d5 string M= NaN
info depth 5 seldepth 7 time 2266 nodes 611 score cp 6 tbhits 0 nps 270 pv d2d4 d7d5 c2c4 c7c6 c4d5 c6d5 string M= NaN
info depth 5 seldepth 8 time 2807 nodes 847 score cp 6 tbhits 0 nps 302 pv d2d4 d7d5 c2c4 c7c6 c4d5 c6d5 c1f4 string M= NaN
info depth 5 seldepth 9 time 3178 nodes 1000 score cp 6 tbhits 0 nps 315 pv d2d4 d7d5 c2c4 c7c6 c4d5 c6d5 c1f4 a7a6 string M= NaN
bestmove d2d4
|=========================================================|
| Ceres - A Monte Carlo Tree Search Chess Engine |
| |
| (c) 2020- David Elliott and the Ceres Authors |
| Use help to list available commands. |
| |
| Version 1.00 [git:20f07b0] |
| Runtime .NET 8.0.11 and CUDA 12.70 |
|=========================================================|
Ceres user settings loaded from C:\eng\Ceres_V1.0.1\Ceres.json
Entering UCI command processing mode.
setoption name WeightsFile value C1-640-25
uci info Network evaluation configured to use: <NNEvaluatorDef Network=Ceres:C1-640-25 Device=GPU:0 >
go
Loaded network weights: C1-640-25
Starting ONNX runtime against C1-640-25.onnx from C1-640-25.onnx with GPU 0 using (CUDA FP16)
0,787 sec TC: ONNX InferenceSession create on model of size 370.359.758 bytes mdiff: 0,00 memTot: 0,00 CPU: 1,08 secs, GC 0 0 0 WS: 0,78 WSPeak: 1,10
Starting ONNX runtime against C1-640-25.onnx from C1-640-25.onnx with GPU 0 using (CUDA FP16)
0,681 sec TC: ONNX InferenceSession create on model of size 370.359.758 bytes mdiff: 0,00 memTot: 0,00 CPU: 1,05 secs, GC 0 0 0 WS: 1,50 WSPeak: 1,82
Unsupported time control in UCI go command go
isready
readyok
go nodes 1000
info depth 1 seldepth 1 time 104 nodes 1 score cp 15 tbhits 0 nps 10 pv e2e4 string M= NaN
info depth 4 seldepth 8 time 616 nodes 120 score cp 8 tbhits 0 nps 195 pv e2e4 e7e5 g1f3 b8c6 f1c4 f8c5 d2d3 string M= NaN
info depth 5 seldepth 9 time 1160 nodes 606 score cp 7 tbhits 0 nps 522 pv e2e4 e7e5 g1f3 b8c6 f1c4 f8c5 d2d3 d7d6 string M= NaN
info depth 6 seldepth 9 time 1520 nodes 1000 score cp 6 tbhits 0 nps 658 pv e2e4 e7e5 g1f3 b8c6 f1c4 f8c5 d2d3 d7d6 string M= NaN
bestmove e2e4Powered by mwForum 2.29.3 © 1999-2014 Markus Wichitill