AlphaZero-style Go engine in C++17/CUDA: HMGS-DP search + 8-expert CA-MOE (62M params), GTP, full training stack. Research prototype; benchmarks and known limits documented.

alphazero cpp17 cuda deep-learning game-ai go go-ai gtp libtorch mcts mixture-of-experts reinforcement-learning research-project self-play uct
2 Open Issues Need Help Last updated: Oct 25, 2026

Open Issues Need Help

View All on GitHub

AlphaZero-style Go engine in C++17/CUDA: HMGS-DP search + 8-expert CA-MOE (62M params), GTP, full training stack. Research prototype; benchmarks and known limits documented.

C++
#alphazero#cpp17#cuda#deep-learning#game-ai#go#go-ai#gtp#libtorch#mcts#mixture-of-experts#reinforcement-learning#research-project#self-play#uct

AlphaZero-style Go engine in C++17/CUDA: HMGS-DP search + 8-expert CA-MOE (62M params), GTP, full training stack. Research prototype; benchmarks and known limits documented.

C++
#alphazero#cpp17#cuda#deep-learning#game-ai#go#go-ai#gtp#libtorch#mcts#mixture-of-experts#reinforcement-learning#research-project#self-play#uct