Texel 1.11

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

Moderator: Ras

petero2
Posts: 725
Joined: Mon Apr 19, 2010 7:07 pm
Location: Sweden
Full name: Peter Osterlund

Texel 1.11

Post by petero2 »

Version 1.11 of my chess engine Texel is now available for download.

This version is around 60 elo stronger than 1.10 at hyper-bullet self play.

The following pre-compiled binaries are included in the download:
  • texel64-avx512.exe
    For Windows 7 or later x86-64 systems using AVX-512, VNNI, POPCOUNT and BMI2 instructions.
  • texel64-avx2-bmi.exe:
    For Windows 7 or later x86-64 systems using AVX2, POPCOUNT and BMI2 instructions.
  • texel64-avx2-pop.exe:
    For Windows 7 or later x86-64 systems using AVX2 and POPCOUNT instructions.
  • texel64-avx2.exe:
    For Windows 7 or later x86-64 systems using AVX2 instructions.
  • texel64-ssse3.exe
    For Windows x86-64 systems with SSSE3 support.
  • texel64-old.exe
    For Windows x86-64 systems.
  • texel64cl.exe:
    Cluster version of texel64-avx2-pop.exe.
    Requires Microsoft MPI to be installed.
  • texel-arm64-dot:
    For the ARMv8-A 64-bit architecture with dot product CPU instruction.
    Should work on many modern Android devices.
  • texel-arm64:
    For the ARMv8-A 64-bit architecture.
    Should work on most modern Android devices.
  • texel64:
    For Linux x86-64 systems with AVX2 and POPCOUNT support.
In most cases the first version in the list that works on your computer is the fastest.

There are no pre-built 32-bit binaries, but 32-bit is still supported in the source code and build system.

List of changes:

Neural network improvements:
- Fewer non-incremental updates of the first layer state.
- Implement quantization aware training.
- Better GPU utilization by using separate data loading threads.
- Efficiently handle training data that is too large to fit in RAM.
- Train a new network based on new training data evaluated by Texel 1.10.
- Optionally use AVX-512 instructions.

Other:
- Re-calibrate UCI_Elo for NNUE.
- Fix a null-move verification problem.
- Update GoogleTest code to version 1.14 to fix warnings when using new CMake versions.
- Add state to the Square class, i.e. make it contain the square number.
Damir
Posts: 2885
Joined: Mon Feb 11, 2008 3:53 pm
Location: Denmark
Full name: Damir Desevac

Re: Texel 1.11

Post by Damir »

Thanks Peter. :) :D
Frank Quisinsky
Posts: 7044
Joined: Wed Nov 18, 2009 7:16 pm
Location: Gutweiler, Germany
Full name: Frank Quisinsky

Re: Texel 1.11

Post by Frank Quisinsky »

Hi Peter,

thanks for Texel 1.11 NN (one of my favourites, no secret).

After the first 90 games the move average for draws is great!
Average of 75 moves (draws, wins, loses) after 90 games.
Of course it could be random after 90 games.

Shitty move averages, have made me unpopular for some time now. I know that but its really important.
Not very for blitz rules, only for longer time controls and when chess players replay comp-comp games and of course some of the statistics.

BY THE WAY:
How do you develop your opening book?
Do you look at all the A00-E99 codes, if so it might be interesting to offer the *.pgn file!
Or do you look more at which of the lines are good for Texel ... in this case the *.pgn file should be secret ... think so.

Thanks again and have a nice start in the year 2024!

Best
Frank
User avatar
Dariusz
Posts: 379
Joined: Sat Jun 13, 2015 10:08 am
Location: Poland
Full name: Dariusz Domagała

Re: Texel 1.11

Post by Dariusz »

Thank you very much, Peter :)
Regards, Darius
https://chessengeria.eu
petero2
Posts: 725
Joined: Mon Apr 19, 2010 7:07 pm
Location: Sweden
Full name: Peter Osterlund

Re: Texel 1.11

Post by petero2 »

Hi Frank,

About the opening book. There is no pgn file that is used to generate the book.

I have written my own program that uses Texel to analyze a large number of positions and then creates an opening book from the analyzed positions. The program uses a technique called drop-out expansion, which is explained briefly here: http://www.fierz.ch/strategy4.htm

My program generates a database of analyzed positions, in a non-standard format that is only understood by my program. I don't consider this file secret, but I have not released it since I don't think it would be useful for anyone but me. My program can currently only be compiled in Linux and it is not user-friendly. It looks like this:

Image

I have not updated the polyglot book in the Texel release in a long time, but will perhaps do that in a coming release, but I should extend the database with analysis using Texel NN first, since it is significantly stronger than the old Texel versions I used when building the current book database.
Frank Quisinsky
Posts: 7044
Joined: Wed Nov 18, 2009 7:16 pm
Location: Gutweiler, Germany
Full name: Frank Quisinsky

Re: Texel 1.11

Post by Frank Quisinsky »

I thought you had some sort of strategy behind it. I'll have a look at the site, link you gave.

I'm so unmotivated on holiday, sitting around with annoying people, more like my sister. I also have some ideas for a new book, but it always takes too long and that puts me off. So I thought I'd ask Peter, I'm sure he's doing his own things, to give me some inspiration for a new project.

Thats my main problem. My old FEOBOS project is out of time (created by a team of engines with classical eval). With a new team of NN engines, maybe 10-20, I can create a stronger book today. Should do that before I start the next bigger tourney with longer time-controls. So, I have a bit to do in free time.