My strongest NNUE net (yet)

[cdani]

Problably new best net, the Gekkehenker one:

40+0.3

Code: Select all

   # PLAYER       : RATING  ERROR   POINTS  PLAYED    (%)
   1 stnnuegk1    : 2866.9    8.7    268.5     506   53.1%
   2 stnnuecl1    : 2845.1    8.7    237.5     506   46.9%

Link to the net:
https://drive.google.com/file/d/1Xgojsp ... 9Szo7/view

Congatulations Gekkehenker!

[cdani]

Gekkehenker comments:
TB6 + draws included in the trainging data.
Almost 1 billion depth 8 positions with TB6.
gensfen depth 8 loop 1000000000 random_max_ply 14 random_move _count 7 write_min_ply 14 output_file_name 1bdepth8.bin

Command to learn:
learn targetdir traindir loop 100 batchsize 1000000 eta 1.0 lambda 1.0 eval_limit 32000 nn_batch_size 1000 newbob_decay 0.5 eval_save_interval 250000000 loss_output_interval 10000000 mirror_percentage 50 validation_set_file_name val\d12.bin

Validation data: Only used 500k D12 sfens

[cdani]

New test at longer time control 200 + 0.5

Code: Select all

   # PLAYER       : RATING  ERROR   POINTS  PLAYED    (%)
   1 stnnuegk1    : 2869.9    7.9    234.0     434   53.9%
   2 stnnuecl1    : 2842.1    7.9    200.0     434   46.1%

[lkaufman]

New test at longer time control 200 + 0.5

Code: Select all

   # PLAYER       : RATING  ERROR   POINTS  PLAYED    (%)
   1 stnnuegk1    : 2869.9    7.9    234.0     434   53.9%
   2 stnnuecl1    : 2842.1    7.9    200.0     434   46.1%

I got a result for the new net (using 60 for slowmover) vs SF 11 of 2 wins, 1 loss, 17 draws at 2' + 1" on seven threads. Small sample, but at least they are in the same ballpark.

[pohl4711]

New test at longer time control 200 + 0.5

Code: Select all

   # PLAYER       : RATING  ERROR   POINTS  PLAYED    (%)
   1 stnnuegk1    : 2869.9    7.9    234.0     434   53.9%
   2 stnnuecl1    : 2842.1    7.9    200.0     434   46.1%

I got a result for the new net (using 60 for slowmover) vs SF 11 of 2 wins, 1 loss, 17 draws at 2' + 1" on seven threads. Small sample, but at least they are in the same ballpark.

An overnight testrun (230 games) of the gk1-net with slowmover 60 gave 41% vs SF 11 (3'+1'', singlethread) (cl1-net was clearly weaker (37%)). Not bad! Now, i will try some slowmover-settings (50, 60, 70, 80, 90, 100). We have to be sure about the best slowmover-setting for the future, here.

[Damir]

Sorry but I can not support this engine in any way. It is not user friendly. When I try and load it in Infinity GUI it stops responding and crashes after few moves when still in book..

[cdani]

Problably new best net, also from Gekkehenker:

Code: Select all

40+0.3 
   # PLAYER       : RATING  ERROR   POINTS  PLAYED    (%)
   1 stnnuegk2    : 2860.6    3.9    930.5    1814   51.3%
   2 stnnuegk1    : 2851.4    3.9    883.5    1814   48.7%

Here the net:
https://cdn.discordapp.com/attachments/ ... 24_june.7z

Thanks to Gekkehenker and also to the community that is helping to improve this engine bit a bit!

[peter]

Sorry but I can not support this engine in any way. It is not user friendly. When I try and load it in Infinity GUI it stops responding and crashes after few moves when still in book..

Have you tried latest binaries made by Norman?

https://github.com/FireFather/sf-nnue/releases

Popcnt- compiles work fine for me.

[JohnS]

Have you tried latest binaries made by Norman?

https://github.com/FireFather/sf-nnue/releases

Popcnt- compiles work fine for me.

Trying to understand this but not much luck so far.

I run sf-nnue run (halfkp-256x2-32-32) x64-bmi2 from the command line and then used
go depth 20

The engine just stopped after some time without any output, is this expected.

Suggestion - allow nn.bin to be in same directory as the exe, this seems simpler to me for using the engine.

What happens if nn.bin is missing, is it just normal stockfish.

[cdani]

Trying to understand this but not much luck so far.

I run sf-nnue run (halfkp-256x2-32-32) x64-bmi2 from the command line and then used
go depth 20

The engine just stopped after some time without any output, is this expected.

Suggestion - allow nn.bin to be in same directory as the exe, this seems simpler to me for using the engine.

What happens if nn.bin is missing, is it just normal stockfish.

Put the net in "eval" folder under the folder of the engine, or change the EvalDir parameter.
Also you need to send an
isready
command
for the engine to load the eval file.