Hier die UCI Engine Seer NNUE mit Sourcecode und TrainingCode:
https://github.com/connormcmonigle/seer-nnueSeer is an original, strong UCI chess engine. Seer relies on a neural network estimating WDL probabilities for position evaluation. Seer's network is trained through a novel retrograde learning approach starting only from 6-man syzygy EGTB WDL values. These initial WDL scores are then iteratively backed up to 32-man chess positions using Seer's search to find continuations from N-man chess positions to N-1-man chess positions (implementation). Seer uses a conventional alpha-beta search combined with "Lazy SMP" (shared transposition table) for multithreading support.
UCI Options
OwnBook (specifies whether or not to use a separate opening book)
BookPath (path to a file containing book positions in a supported format)
Threads (for every thread doubling, a gain of about 70-80 elo can be expected)
Hash (the amount of the memory allocated for the transposition table (actual memory usage will be greater))
Weights (the absolute path to a binary weights file. If the default "EMBEDDED" path is chosen, the embedded weights will be used.)
Features
From scratch neural network training and execution (using OpenMP SIMD directives and SIMD intrinsics) implementation (training scripts use PyTorch for GPU acceleration and can be found here).
Plain magic bitboard move generation with constexpr compile time generated attack tables.
Principal variation search inside an iterative deepening framework
Lockless shared transposition table (using Zobrist hashing)
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
History extensions
Null move pruning
Static null move pruning (reverse futility pruning)
Futility pruning
Late move reductions
Aspiration windows
Forum
CSS-Shop