SSSimpler is a UCI chess engine with NNUE.
Alpha-Beta search with a bunch of tweaks
Neural net, architecture is (768x8hm->1024)x2-pw->(16-dual->32-crelu->1)x(8x8).
All data came from self-play.
Just run make
Yet to be tested, estimated 3650 CCRL blitz
The code is licensed under the GPLv3 license. Full text can be found in LICENSE in the project root.
Jonathan Hallström, author of Pawnocchio, for a ton of helping throughout the development, SSSimpler would be much much weaker without him
Rektbench gang for helping and providing hardware for testing
Bullet NNUE trainer for training all nets
Alpha-Beta Discord server for explaining some concepts