HCha/Stockfish
1
0
Fork 0
mirror of https://github.com/HChaZZY/Stockfish.git synced 2025-12-20 17:16:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add NNUE evaluation

Browse Source 
This patch ports the efficiently updatable neural network (NNUE) evaluation to Stockfish.

Both the NNUE and the classical evaluations are available, and can be used to
assign a value to a position that is later used in alpha-beta (PVS) search to find the
best move. The classical evaluation computes this value as a function of various chess
concepts, handcrafted by experts, tested and tuned using fishtest. The NNUE evaluation
computes this value with a neural network based on basic inputs. The network is optimized
and trained on the evalutions of millions of positions at moderate search depth.

The NNUE evaluation was first introduced in shogi, and ported to Stockfish afterward.
It can be evaluated efficiently on CPUs, and exploits the fact that only parts
of the neural network need to be updated after a typical chess move.
[The nodchip repository](https://github.com/nodchip/Stockfish) provides additional
tools to train and develop the NNUE networks.

This patch is the result of contributions of various authors, from various communities,
including: nodchip, ynasu87, yaneurao (initial port and NNUE authors), domschl, FireFather,
rqs, xXH4CKST3RXx, tttak, zz4032, joergoster, mstembera, nguyenpham, erbsenzaehler,
dorzechowski, and vondele.

This new evaluation needed various changes to fishtest and the corresponding infrastructure,
for which tomtor, ppigazzini, noobpwnftw, daylen, and vondele are gratefully acknowledged.

The first networks have been provided by gekkehenker and sergiovieri, with the latter
net (nn-97f742aaefcd.nnue) being the current default.

The evaluation function can be selected at run time with the `Use NNUE` (true/false) UCI option,
provided the `EvalFile` option points the the network file (depending on the GUI, with full path).

The performance of the NNUE evaluation relative to the classical evaluation depends somewhat on
the hardware, and is expected to improve quickly, but is currently on > 80 Elo on fishtest:

60000 @ 10+0.1 th 1
https://tests.stockfishchess.org/tests/view/5f28fe6ea5abc164f05e4c4c
ELO: 92.77 +-2.1 (95%) LOS: 100.0%
Total: 60000 W: 24193 L: 8543 D: 27264
Ptnml(0-2): 609, 3850, 9708, 10948, 4885

40000 @ 20+0.2 th 8
https://tests.stockfishchess.org/tests/view/5f290229a5abc164f05e4c58
ELO: 89.47 +-2.0 (95%) LOS: 100.0%
Total: 40000 W: 12756 L: 2677 D: 24567
Ptnml(0-2): 74, 1583, 8550, 7776, 2017

At the same time, the impact on the classical evaluation remains minimal, causing no significant
regression:

sprt @ 10+0.1 th 1
https://tests.stockfishchess.org/tests/view/5f2906a2a5abc164f05e4c5b
LLR: 2.94 (-2.94,2.94) {-6.00,-4.00}
Total: 34936 W: 6502 L: 6825 D: 21609
Ptnml(0-2): 571, 4082, 8434, 3861, 520

sprt @ 60+0.6 th 1
https://tests.stockfishchess.org/tests/view/5f2906cfa5abc164f05e4c5d
LLR: 2.93 (-2.94,2.94) {-6.00,-4.00}
Total: 10088 W: 1232 L: 1265 D: 7591
Ptnml(0-2): 49, 914, 3170, 843, 68

The needed networks can be found at https://tests.stockfishchess.org/nns
It is recommended to use the default one as indicated by the `EvalFile` UCI option.

Guidelines for testing new nets can be found at
https://github.com/glinscott/fishtest/wiki/Creating-my-first-test#nnue-net-tests

Integration has been discussed in various issues:
https://github.com/official-stockfish/Stockfish/issues/2823
https://github.com/official-stockfish/Stockfish/issues/2728

The integration branch will be closed after the merge:
https://github.com/official-stockfish/Stockfish/pull/2825
https://github.com/official-stockfish/Stockfish/tree/nnue-player-wip

closes https://github.com/official-stockfish/Stockfish/pull/2912

This will be an exciting time for computer chess, looking forward to seeing the evolution of
this approach.

Bench: 4746616
This commit is contained in:
nodchip
2020-08-05 17:11:15 +02:00
committed by Joost VandeVondele
parent 9587eeeb5e
commit 84f3e86790
59 changed files with 2474 additions and 245 deletions
Download Patch File Download Diff File Expand all files Collapse all files

129
src/types.h
View File

@@ -1,8 +1,6 @@
/*
Stockfish, a UCI chess playing engine derived from Glaurung 2.1
Copyright (C) 2004-2008 Tord Romstad (Glaurung author)
Copyright (C) 2008-2015 Marco Costalba, Joona Kiiski, Tord Romstad
Copyright (C) 2015-2020 Marco Costalba, Joona Kiiski, Gary Linscott, Tord Romstad
Copyright (C) 2004-2020 The Stockfish developers (see AUTHORS file)
Stockfish is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
@@ -203,6 +201,22 @@ enum Piece {
PIECE_NB = 16
};
// An ID used to track the pieces. Max. 32 pieces on board.
enum PieceId {
PIECE_ID_ZERO = 0,
PIECE_ID_KING = 30,
PIECE_ID_WKING = 30,
PIECE_ID_BKING = 31,
PIECE_ID_NONE = 32
};
inline PieceId operator++(PieceId& d, int) {
PieceId x = d;
d = PieceId(int(d) + 1);
return x;
}
constexpr Value PieceValue[PHASE_NB][PIECE_NB] = {
{ VALUE_ZERO, PawnValueMg, KnightValueMg, BishopValueMg, RookValueMg, QueenValueMg, VALUE_ZERO, VALUE_ZERO,
VALUE_ZERO, PawnValueMg, KnightValueMg, BishopValueMg, RookValueMg, QueenValueMg, VALUE_ZERO, VALUE_ZERO },
@@ -232,7 +246,8 @@ enum Square : int {
SQ_A8, SQ_B8, SQ_C8, SQ_D8, SQ_E8, SQ_F8, SQ_G8, SQ_H8,
SQ_NONE,
SQUARE_NB = 64
SQUARE_ZERO = 0,
SQUARE_NB = 64
};
enum Direction : int {
@@ -255,6 +270,94 @@ enum Rank : int {
RANK_1, RANK_2, RANK_3, RANK_4, RANK_5, RANK_6, RANK_7, RANK_8, RANK_NB
};
// unique number for each piece type on each square
enum PieceSquare : uint32_t {
PS_NONE = 0,
PS_W_PAWN = 1,
PS_B_PAWN = 1 * SQUARE_NB + 1,
PS_W_KNIGHT = 2 * SQUARE_NB + 1,
PS_B_KNIGHT = 3 * SQUARE_NB + 1,
PS_W_BISHOP = 4 * SQUARE_NB + 1,
PS_B_BISHOP = 5 * SQUARE_NB + 1,
PS_W_ROOK = 6 * SQUARE_NB + 1,
PS_B_ROOK = 7 * SQUARE_NB + 1,
PS_W_QUEEN = 8 * SQUARE_NB + 1,
PS_B_QUEEN = 9 * SQUARE_NB + 1,
PS_W_KING = 10 * SQUARE_NB + 1,
PS_END = PS_W_KING, // pieces without kings (pawns included)
PS_B_KING = 11 * SQUARE_NB + 1,
PS_END2 = 12 * SQUARE_NB + 1
};
struct ExtPieceSquare {
PieceSquare from[COLOR_NB];
};
// Array for finding the PieceSquare corresponding to the piece on the board
extern ExtPieceSquare kpp_board_index[PIECE_NB];
constexpr bool is_ok(PieceId pid);
constexpr Square rotate180(Square sq);
// Structure holding which tracked piece (PieceId) is where (PieceSquare)
class EvalList {
public:
// Max. number of pieces without kings is 30 but must be a multiple of 4 in AVX2
static const int MAX_LENGTH = 32;
// Array that holds the piece id for the pieces on the board
PieceId piece_id_list[SQUARE_NB];
// List of pieces, separate from White and Black POV
PieceSquare* piece_list_fw() const { return const_cast<PieceSquare*>(pieceListFw); }
PieceSquare* piece_list_fb() const { return const_cast<PieceSquare*>(pieceListFb); }
// Place the piece pc with piece_id on the square sq on the board
void put_piece(PieceId piece_id, Square sq, Piece pc)
{
assert(is_ok(piece_id));
if (pc != NO_PIECE)
{
pieceListFw[piece_id] = PieceSquare(kpp_board_index[pc].from[WHITE] + sq);
pieceListFb[piece_id] = PieceSquare(kpp_board_index[pc].from[BLACK] + rotate180(sq));
piece_id_list[sq] = piece_id;
}
else
{
pieceListFw[piece_id] = PS_NONE;
pieceListFb[piece_id] = PS_NONE;
piece_id_list[sq] = piece_id;
}
}
// Convert the specified piece_id piece to ExtPieceSquare type and return it
ExtPieceSquare piece_with_id(PieceId piece_id) const
{
ExtPieceSquare eps;
eps.from[WHITE] = pieceListFw[piece_id];
eps.from[BLACK] = pieceListFb[piece_id];
return eps;
}
private:
PieceSquare pieceListFw[MAX_LENGTH];
PieceSquare pieceListFb[MAX_LENGTH];
};
// For differential evaluation of pieces that changed since last turn
struct DirtyPiece {
// Number of changed pieces
int dirty_num;
// The ids of changed pieces, max. 2 pieces can change in one move
PieceId pieceId[2];
// What changed from the piece with that piece number
ExtPieceSquare old_piece[2];
ExtPieceSquare new_piece[2];
};
/// Score enum stores a middlegame and an endgame value in a single integer (enum).
/// The least significant 16 bits are used to store the middlegame value and the
@@ -280,10 +383,10 @@ inline Value mg_value(Score s) {
}
#define ENABLE_BASE_OPERATORS_ON(T) \
constexpr T operator+(T d1, int d2) { return T(int(d1) + d2); } \
constexpr T operator-(T d1, int d2) { return T(int(d1) - d2); } \
constexpr T operator+(T d1, int d2) { return T(int(d1) + d2); } \
constexpr T operator-(T d1, int d2) { return T(int(d1) - d2); } \
constexpr T operator-(T d) { return T(-int(d)); } \
inline T& operator+=(T& d1, int d2) { return d1 = d1 + d2; } \
inline T& operator+=(T& d1, int d2) { return d1 = d1 + d2; } \
inline T& operator-=(T& d1, int d2) { return d1 = d1 - d2; }
#define ENABLE_INCR_OPERATORS_ON(T) \
@@ -302,6 +405,9 @@ inline T& operator/=(T& d, int i) { return d = T(int(d) / i); }
ENABLE_FULL_OPERATORS_ON(Value)
ENABLE_FULL_OPERATORS_ON(Direction)
ENABLE_INCR_OPERATORS_ON(Piece)
ENABLE_INCR_OPERATORS_ON(PieceSquare)
ENABLE_INCR_OPERATORS_ON(PieceId)
ENABLE_INCR_OPERATORS_ON(PieceType)
ENABLE_INCR_OPERATORS_ON(Square)
ENABLE_INCR_OPERATORS_ON(File)
@@ -390,6 +496,10 @@ inline Color color_of(Piece pc) {
return Color(pc >> 3);
}
constexpr bool is_ok(PieceId pid) {
return pid < PIECE_ID_NONE;
}
constexpr bool is_ok(Square s) {
return s >= SQ_A1 && s <= SQ_H8;
}
@@ -426,6 +536,11 @@ constexpr Square to_sq(Move m) {
return Square(m & 0x3F);
}
// Return relative square when turning the board 180 degrees
constexpr Square rotate180(Square sq) {
return (Square)(sq ^ 0x3F);
}
constexpr int from_to(Move m) {
return m & 0xFFF;
}