Chess2u
Would you like to react to this message? Create an account in a few clicks or log in to continue.

Chess2uLog in

CFish

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
I'm missing this file: nn-202008141035.nnue
Can some send me a link just for that file only

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Waldemar wrote:
Hybrid better?


Yes.

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Match: Cfish Dev Hybrid vs Pure Scaling.
Tournament: Gauntlet
Binaries: Cfish.exe (Dev)
Engine Options: For Cfish Dev Hybrid = Default  |  For Cfish Dev Pure = Pure instead of Hybrid in engine config.
Hardware: AMD Ryzen 3700X
Time control: 10+0.1, 20+0.2, 40+0.4, 80+0.8, 300+3   (Speed for CFish NNUE  ~ 1250 Knps)
Book: 8moves_v3.pgn
Tablebases: None
Adjudication: +-8 Win
Software: Cutechess-Cli GUI
Comments: Difference between Hybrid vs Pure remains more or less the same irrespective of time controls.                                                                                    

Code:


Time control = 10 + 0.1

  # PLAYER               :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)     W     D     L  D(%)
   1 Cfish-Dev Hybrid     :    16.0    3.2  5226.0   10000  52.3     100  1345  7762   893  77.6
   2 Cfish-Dev Pure       :     0.0   ----  4774.0   10000  47.7     ---   893  7762  1345  77.6

White advantage = 30.85 +/- 1.65
Draw rate (equal opponents) = 79.40 % +/- 0.42

                                                                          
Time control = 20 + 0.2

   # PLAYER               :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)     W     D     L  D(%)
   1 Cfish-Dev Hybrid     :    13.5    2.8  5190.5   10000  51.9     100  1053  8275   672  82.8
   2 Cfish-Dev Pure       :     0.0   ----  4809.5   10000  48.1     ---   672  8275  1053  82.8

White advantage = 29.31 +/- 1.44
Draw rate (equal opponents) = 84.91 % +/- 0.39

                                                                        
 Time control = 40 + 0.4

   # PLAYER               :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W     D    L  D(%)
   1 Cfish-Dev Hybrid     :    14.3    3.7  2601.0    5000  52.0     100  464  4274  262  85.5
   2 Cfish-Dev Pure       :     0.0   ----  2399.0    5000  48.0     ---  262  4274  464  85.5

White advantage = 29.42 +/- 1.86
Draw rate (equal opponents) = 88.27 % +/- 0.51

Time control = 80 + 0.8

   # PLAYER               :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W     D    L  D(%)
   1 Cfish-Dev Hybrid     :    14.8    4.2  1823.5    3500  52.1     100  298  3051  151  87.2
   2 Cfish-Dev Pure       :     0.0   ----  1676.5    3500  47.9     ---  151  3051  298  87.2

White advantage = 29.26 +/- 2.06
Draw rate (equal opponents) = 90.48 % +/- 0.57


Time control = 300 + 3

   # PLAYER               :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W    D    L  D(%)
   1 Cfish-Dev Hybrid     :    13.7    6.3   519.5    1000  52.0     100   65  909   26  90.9
   2 Cfish-Dev Pure       :     0.0   ----   480.5    1000  48.0     ---   26  909   65  90.9

White advantage = 24.26 +/- 3.27
Draw rate (equal opponents) = 94.49 % +/- 0.90

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Debug Log File???

Some guys and I have seen freezes, crashes and doesn't start problems and other things.

Why CFish has no debug log file like Stockfish?
I think this is maybe the biggest mistake of CFish.

Bugs and problems could be detected much faster and much easier and solved much faster.

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Mario wrote:
Waldemar wrote:
Hybrid better?


Yes.

I use cfish 230920, nnue classic + large pages, for the great speed. Ty, jorge.

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Cfish: Updated October, 7 2020.
For Windows, Android 64bit and Linux, NO ANDROID 32BIT OR MACOS!

- Changelog: Updated to "Pawn Tuning".

HUGE BINARIES ASSORTMENT FOR 32BIT SYSTEMS!

Remember to place always the nnue file in the package together with the program, exception made for the Android release.

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
lisandr wrote:


This engine is very strong playing on Android platform. It is very fast, as it manages to beat StockFish12 at + 300 Kn/s playing with both engines with 4 cores. In my first test Cfish achieves a performance of +46 Elo.  eek

Amazing!!! clap

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
denochss wrote:
lisandr wrote:


This engine is very strong playing on Android platform. It is very fast, as it manages to beat StockFish12 at + 300 Kn/s playing with both engines with 4 cores. In my first test Cfish achieves a performance of +46 Elo.  eek

Amazing!!! clap

Thank u! Do you know where I can download the nnue network nn-202008141035.nnue?

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Thank u! Do you know where I can download the nnue network nn-202008141035.nnue?[/quote]

Go to this site..

https://www.comp.nus.edu.sg/~sergio-v/nnue/

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Thanks, Zippo

descriptionCFish - Page 7 Emptynnue

more_horiz

https://www.comp.nus.edu.sg/~sergio-v/nnue/

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Сompilation from ChessMan

Cfish 071020 x32-x64


Updated to "Pawn Tuning".
syzygy1 committed 2 days ago

Remove some more permutation instructions.1
syzygy1 committed 3 days ago

Remove a few permutations.
syzygy1 committed 3 days ago

Merge pull request #168 from gsobala/patch-2
syzygy1 committed 3 days ago

Type error (again) in nnue.c
gsobala committed 3 days ago

MMX sparse multiplication tiling.
syzygy1 committed 4 days ago


Windows: https://pixeldrain.com/u/nwMBH59r
Linux BMI2: https://pixeldrain.com/u/ELS2hZVi

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Linux
AVX2 + SSE4.1 + SSE2_no_popcnt https://pixeldrain.com/u/fQW2WQKY

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
3 new Stockfish compiles;)

How long does it take for the new Cfish?

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Mario wrote:
3 new Stockfish compiles;)

How long does it take for the new Cfish?



Hi Mario.. Try this
Thx to Chessman for the compilation 👍

Updated to "Pawn Tuning".
syzygy1 committed 2 days ago

Remove some more permutation instructions.1
syzygy1 committed 3 days ago


Remove a few permutations.
syzygy1 committed 3 days ago

Merge pull request #168 from gsobala/patch-2
syzygy1 committed 3 days ago

Type error (again) in nnue.c
gsobala committed 3 days ago

MMX sparse multiplication tiling.
syzygy1 committed 4 days ago

Windows: https://pixeldrain.com/u/nwMBH59r
Linux BMI2: https://pixeldrain.com/u/ELS2hZVi

descriptionCFish - Page 7 Emptythis version is

more_horiz



from 7 oct 2020???
Right???\


Zippo wrote:
Mario wrote:
3 new Stockfish compiles;)

How long does it take for the new Cfish?



Hi Mario.. Try this
Thx to Chessman for the compilation 👍

Updated to "Pawn Tuning".
syzygy1 committed 2 days ago

Remove some more permutation instructions.1
syzygy1 committed 3 days ago


Remove a few permutations.
syzygy1 committed 3 days ago

Merge pull request #168 from gsobala/patch-2
syzygy1 committed 3 days ago

Type error (again) in nnue.c
gsobala committed 3 days ago

MMX sparse multiplication tiling.
syzygy1 committed 4 days ago

Windows: https://pixeldrain.com/u/nwMBH59r
Linux BMI2: https://pixeldrain.com/u/ELS2hZVi

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Yeaa.. u right


Powerchess wrote:



from 7 oct 2020???
Right???\


Zippo wrote:
Mario wrote:
3 new Stockfish compiles;)

How long does it take for the new Cfish?



Hi Mario.. Try this
Thx to Chessman for the compilation 👍

Updated to "Pawn Tuning".
syzygy1 committed 2 days ago

Remove some more permutation instructions.1
syzygy1 committed 3 days ago


Remove a few permutations.
syzygy1 committed 3 days ago

Merge pull request #168 from gsobala/patch-2
syzygy1 committed 3 days ago

Type error (again) in nnue.c
gsobala committed 3 days ago

MMX sparse multiplication tiling.
syzygy1 committed 4 days ago

Windows: https://pixeldrain.com/u/nwMBH59r
Linux BMI2: https://pixeldrain.com/u/ELS2hZVi

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Сompilation from ChessMan

Cfish 161020 x32-x64


Updated to "Scale factor tweak". …
syzygy1 committed 16 hours ago

Clear up NNUE transform().
syzygy1 committed 5 days ago

Remove non-sparse multiplication code from nnue.c. …
syzygy1 committed 6 days ago

Tiny SSE2 improvements.
syzygy1 committed 8 days ago

https://pixeldrain.com/u/AsdtTEic

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
lisandr wrote:
Сompilation from ChessMan

Cfish 161020 x32-x64


Updated to "Scale factor tweak". …
syzygy1 committed 16 hours ago

Clear up NNUE transform().
syzygy1 committed 5 days ago

Remove non-sparse multiplication code from nnue.c. …
syzygy1 committed 6 days ago

Tiny SSE2 improvements.
syzygy1 committed 8 days ago

https://pixeldrain.com/u/AsdtTEic



I see we have here the new net: "nn-04cf2b4ed1da.nnue" and the "Scale factor tweak" but where is the "RookOnQueenFile Removal Elo +4.33"?

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz

Code:

Result:
---------------------------------------------------------------------------------------
  #  name                       games    wins   draws  losses   score    los%  elo+/-
  1. BrainLearn 12.1              500      24     454      22   251.0    61.6     [B][COLOR="Red"]1.4[/COLOR][/B]
  2. Cfish 161020 64 BMI2 NUMA    500      22     454      24   249.0    38.4    -1.4

Tech:
---------------------------------------------------------------------------------------

Tech (average nodes, depths, time/m per move, others per game):
  #  name                         nodes/m         NPS  depth/m   time/m    moves     time
  1. BrainLearn 12.1                2054K      681613     26.3      3.0     44.1    133.0
  2. Cfish 161020 64 BMI2 NUMA      2361K      782510     26.8      3.0     44.2    133.2


Match: BrainLearn 12.1 vs Cfish 161020
Binaries: BrainLearn12.1-x86-64-bmi2.exe, Cfish_x64_161020_E_BMI2_mingw_10.exe
Hash: 128MB
Threads: 1, concurrency=40
Hardware: 2x Intel(R) Xeon(R) CPU E5-2678v3 @ 2.80GHz
Time control: 2'+1"
Book: Hert_lowdraws_250.pgn
Tablebases: 6-men Syzygy for GUI
Adjudication: -draw movenumber=30 movecount=5 score=9, -resign movecount=2 score=999
Software: Banksia GUI 0.37

https://github.com/amchess/BrainLearn/releases/tag/12.1 -

https://github.com/amchess/ShashChess/releases/15.1

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
CF_EXT 181020 x32-x64

Сompilation from ChessMan

https://pixeldrain.com/u/4gSpp2JE

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
lisandr wrote:
CF_EXT 181020 x32-x64

Сompilation from ChessMan

https://pixeldrain.com/u/4gSpp2JE


It would be interesting and good to see the improvements.
Because some other guys and I often saw in the last time that compiles are missed.

See also:
Where is the Stockfish compile from 20-10-11??
https://github.com/official-stockfish/Stockfish/issues/3182
When I remember correctly Stockfish developers had missed this compile and after 3 new compiles some days or a week later, someone wrote this thread and then they have created this compile.
How is it possible that they often forget to create the compiles from passed green tests?

See also a new one some days ago missed:
20-10-15 Viz fhCntRedLTCQ2 diff
LLR: 2.94 (-2.94,2.94) {0.25,1.25}
Total: 59136 W: 2773 L: 2564 D: 53799
Ptnml(0-2): 30, 2117, 25078, 2300, 43
sprt @ 60+0.6 th 1
Take 2
https://tests.stockfishchess.org/tests/view/5f884dbfdcdad978fe8c527a
https://github.com/Vizvezdenec/Stockfish/compare/288a604411...96408fc440

And another one:
20-10-16 Fis Hid1Bias diff
LLR: 2.96 (-2.94,2.94) {0.25,1.25}
Total: 81272 W: 3948 L: 3682 D: 73642
Ptnml(0-2): 64, 3194, 33856, 3456, 66
sprt @ 60+0.6 th 1
LTC Result of https://tests.stockfishchess.org/tests/view/5f875213dcdad978fe8c5211 (If this passes I will write out the actual .nnue file.)
https://tests.stockfishchess.org/tests/view/5f89e8efeae8a6e60644d6e7
https://github.com/mstembera/Stockfish/compare/288a604411...432da86721

Why there is no link which shows: xxx test passed and a compile is needed.
It wouldn't be possible to forget it.

descriptionCFish - Page 7 EmptyBrainLearn 12.1

more_horiz
Where from can i download Brainlearn 12.1 or latest?
Thx



lisandr wrote:

Code:

Result:
---------------------------------------------------------------------------------------
  #  name                       games    wins   draws  losses   score    los%  elo+/-
  1. BrainLearn 12.1              500      24     454      22   251.0    61.6     [B][COLOR="Red"]1.4[/COLOR][/B]
  2. Cfish 161020 64 BMI2 NUMA    500      22     454      24   249.0    38.4    -1.4

Tech:
---------------------------------------------------------------------------------------

Tech (average nodes, depths, time/m per move, others per game):
  #  name                         nodes/m         NPS  depth/m   time/m    moves     time
  1. BrainLearn 12.1                2054K      681613     26.3      3.0     44.1    133.0
  2. Cfish 161020 64 BMI2 NUMA      2361K      782510     26.8      3.0     44.2    133.2


Match: BrainLearn 12.1 vs Cfish 161020
Binaries: BrainLearn12.1-x86-64-bmi2.exe, Cfish_x64_161020_E_BMI2_mingw_10.exe
Hash: 128MB
Threads: 1, concurrency=40
Hardware: 2x Intel(R) Xeon(R) CPU E5-2678v3 @ 2.80GHz
Time control: 2'+1"
Book: Hert_lowdraws_250.pgn
Tablebases: 6-men Syzygy for GUI
Adjudication: -draw movenumber=30 movecount=5 score=9, -resign movecount=2 score=999
Software: Banksia GUI 0.37

https://github.com/amchess/BrainLearn/releases/tag/12.1 -

https://github.com/amchess/ShashChess/releases/15.1

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Powerchess wrote:
Where from can i download Brainlearn 12.1 or latest?
Thx


https://github.com/amchess/BrainLearn/releases

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Cfish 221020 NNUE.

https://rapidu.net/2881593611/Cfish-221020-NNUE.7z

descriptionCFish - Page 7 EmptyRe: CFish

more_horiz
Permissions in this forum:
You cannot reply to topics in this forum