8000 GitHub - connormcmonigle/seer-nnue at v1.1
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

connormcmonigle/seer-nnue

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

95 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Seer

Seer is a, from scratch, work in progress, UCI chess engine using NNUE for position evaluation. As an engine, Seer stands on the shoulders of many great (and mostly much stronger) predecessors. In particular, Ethereal, Winter, and Stockfish have proven useful in 723D the development of Seer.

UCI Options

  • Clear Hash
  • Threads (for every thread doubling, a gain of about 70-80 elo can be expected)
  • Hash
  • Weights (The absolute path to a binary weights file. This option must be set.)

Features

  • Completely from scratch NNUE training and execution (using OpenMP SIMD directives) implementation (training scripts use PyTorch for GPU acceleration and can be found in the /train subdirectory)
  • Bitboard move generation using PEXT/PDEP instructions with constexpr compile time generated attack tables.
  • Principal variation search inside an iterative deepening framework
  • Lockless shared transposition table (board state Zobrist hash is incrementally updated)
  • Move Ordering (SEE for captures + Killer Move, Combined Butterfly History, Counter Move History and Follow Up History for quiets)
  • History pruning as well as SEE pruning in QSearch
  • Null move pruning
  • Late move reduction
  • Aspiration windows

Compiling

cd build
cmake ..
make seer

About

UCI chess engine using neural networks for position evaluation

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

0