mirror of
https://github.com/opelly27/Stockfish.git
synced 2026-05-20 15:37:47 +00:00
84f3e86790
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
204 lines
7.1 KiB
C++
204 lines
7.1 KiB
C++
/*
|
|
Stockfish, a UCI chess playing engine derived from Glaurung 2.1
|
|
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
|
|
the Free Software Foundation, either version 3 of the License, or
|
|
(at your option) any later version.
|
|
|
|
Stockfish is distributed in the hope that it will be useful,
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
GNU General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
along with this program. If not, see <http://www.gnu.org/licenses/>.
|
|
*/
|
|
|
|
#include <algorithm>
|
|
#include <bitset>
|
|
|
|
#include "bitboard.h"
|
|
#include "misc.h"
|
|
|
|
uint8_t PopCnt16[1 << 16];
|
|
uint8_t SquareDistance[SQUARE_NB][SQUARE_NB];
|
|
|
|
Bitboard SquareBB[SQUARE_NB];
|
|
Bitboard LineBB[SQUARE_NB][SQUARE_NB];
|
|
Bitboard PseudoAttacks[PIECE_TYPE_NB][SQUARE_NB];
|
|
Bitboard PawnAttacks[COLOR_NB][SQUARE_NB];
|
|
|
|
Magic RookMagics[SQUARE_NB];
|
|
Magic BishopMagics[SQUARE_NB];
|
|
|
|
namespace {
|
|
|
|
Bitboard RookTable[0x19000]; // To store rook attacks
|
|
Bitboard BishopTable[0x1480]; // To store bishop attacks
|
|
|
|
void init_magics(PieceType pt, Bitboard table[], Magic magics[]);
|
|
}
|
|
|
|
|
|
/// Bitboards::pretty() returns an ASCII representation of a bitboard suitable
|
|
/// to be printed to standard output. Useful for debugging.
|
|
|
|
const std::string Bitboards::pretty(Bitboard b) {
|
|
|
|
std::string s = "+---+---+---+---+---+---+---+---+\n";
|
|
|
|
for (Rank r = RANK_8; r >= RANK_1; --r)
|
|
{
|
|
for (File f = FILE_A; f <= FILE_H; ++f)
|
|
s += b & make_square(f, r) ? "| X " : "| ";
|
|
|
|
s += "| " + std::to_string(1 + r) + "\n+---+---+---+---+---+---+---+---+\n";
|
|
}
|
|
s += " a b c d e f g h\n";
|
|
|
|
return s;
|
|
}
|
|
|
|
|
|
/// Bitboards::init() initializes various bitboard tables. It is called at
|
|
/// startup and relies on global objects to be already zero-initialized.
|
|
|
|
void Bitboards::init() {
|
|
|
|
for (unsigned i = 0; i < (1 << 16); ++i)
|
|
PopCnt16[i] = uint8_t(std::bitset<16>(i).count());
|
|
|
|
for (Square s = SQ_A1; s <= SQ_H8; ++s)
|
|
SquareBB[s] = (1ULL << s);
|
|
|
|
for (Square s1 = SQ_A1; s1 <= SQ_H8; ++s1)
|
|
for (Square s2 = SQ_A1; s2 <= SQ_H8; ++s2)
|
|
SquareDistance[s1][s2] = std::max(distance<File>(s1, s2), distance<Rank>(s1, s2));
|
|
|
|
init_magics(ROOK, RookTable, RookMagics);
|
|
init_magics(BISHOP, BishopTable, BishopMagics);
|
|
|
|
for (Square s1 = SQ_A1; s1 <= SQ_H8; ++s1)
|
|
{
|
|
PawnAttacks[WHITE][s1] = pawn_attacks_bb<WHITE>(square_bb(s1));
|
|
PawnAttacks[BLACK][s1] = pawn_attacks_bb<BLACK>(square_bb(s1));
|
|
|
|
for (int step : {-9, -8, -7, -1, 1, 7, 8, 9} )
|
|
PseudoAttacks[KING][s1] |= safe_destination(s1, step);
|
|
|
|
for (int step : {-17, -15, -10, -6, 6, 10, 15, 17} )
|
|
PseudoAttacks[KNIGHT][s1] |= safe_destination(s1, step);
|
|
|
|
PseudoAttacks[QUEEN][s1] = PseudoAttacks[BISHOP][s1] = attacks_bb<BISHOP>(s1, 0);
|
|
PseudoAttacks[QUEEN][s1] |= PseudoAttacks[ ROOK][s1] = attacks_bb< ROOK>(s1, 0);
|
|
|
|
for (PieceType pt : { BISHOP, ROOK })
|
|
for (Square s2 = SQ_A1; s2 <= SQ_H8; ++s2)
|
|
if (PseudoAttacks[pt][s1] & s2)
|
|
LineBB[s1][s2] = (attacks_bb(pt, s1, 0) & attacks_bb(pt, s2, 0)) | s1 | s2;
|
|
}
|
|
}
|
|
|
|
|
|
namespace {
|
|
|
|
Bitboard sliding_attack(PieceType pt, Square sq, Bitboard occupied) {
|
|
|
|
Bitboard attacks = 0;
|
|
Direction RookDirections[4] = {NORTH, SOUTH, EAST, WEST};
|
|
Direction BishopDirections[4] = {NORTH_EAST, SOUTH_EAST, SOUTH_WEST, NORTH_WEST};
|
|
|
|
for(Direction d : (pt == ROOK ? RookDirections : BishopDirections))
|
|
{
|
|
Square s = sq;
|
|
while(safe_destination(s, d) && !(occupied & s))
|
|
attacks |= (s += d);
|
|
}
|
|
|
|
return attacks;
|
|
}
|
|
|
|
|
|
// init_magics() computes all rook and bishop attacks at startup. Magic
|
|
// bitboards are used to look up attacks of sliding pieces. As a reference see
|
|
// www.chessprogramming.org/Magic_Bitboards. In particular, here we use the so
|
|
// called "fancy" approach.
|
|
|
|
void init_magics(PieceType pt, Bitboard table[], Magic magics[]) {
|
|
|
|
// Optimal PRNG seeds to pick the correct magics in the shortest time
|
|
int seeds[][RANK_NB] = { { 8977, 44560, 54343, 38998, 5731, 95205, 104912, 17020 },
|
|
{ 728, 10316, 55013, 32803, 12281, 15100, 16645, 255 } };
|
|
|
|
Bitboard occupancy[4096], reference[4096], edges, b;
|
|
int epoch[4096] = {}, cnt = 0, size = 0;
|
|
|
|
for (Square s = SQ_A1; s <= SQ_H8; ++s)
|
|
{
|
|
// Board edges are not considered in the relevant occupancies
|
|
edges = ((Rank1BB | Rank8BB) & ~rank_bb(s)) | ((FileABB | FileHBB) & ~file_bb(s));
|
|
|
|
// Given a square 's', the mask is the bitboard of sliding attacks from
|
|
// 's' computed on an empty board. The index must be big enough to contain
|
|
// all the attacks for each possible subset of the mask and so is 2 power
|
|
// the number of 1s of the mask. Hence we deduce the size of the shift to
|
|
// apply to the 64 or 32 bits word to get the index.
|
|
Magic& m = magics[s];
|
|
m.mask = sliding_attack(pt, s, 0) & ~edges;
|
|
m.shift = (Is64Bit ? 64 : 32) - popcount(m.mask);
|
|
|
|
// Set the offset for the attacks table of the square. We have individual
|
|
// table sizes for each square with "Fancy Magic Bitboards".
|
|
m.attacks = s == SQ_A1 ? table : magics[s - 1].attacks + size;
|
|
|
|
// Use Carry-Rippler trick to enumerate all subsets of masks[s] and
|
|
// store the corresponding sliding attack bitboard in reference[].
|
|
b = size = 0;
|
|
do {
|
|
occupancy[size] = b;
|
|
reference[size] = sliding_attack(pt, s, b);
|
|
|
|
if (HasPext)
|
|
m.attacks[pext(b, m.mask)] = reference[size];
|
|
|
|
size++;
|
|
b = (b - m.mask) & m.mask;
|
|
} while (b);
|
|
|
|
if (HasPext)
|
|
continue;
|
|
|
|
PRNG rng(seeds[Is64Bit][rank_of(s)]);
|
|
|
|
// Find a magic for square 's' picking up an (almost) random number
|
|
// until we find the one that passes the verification test.
|
|
for (int i = 0; i < size; )
|
|
{
|
|
for (m.magic = 0; popcount((m.magic * m.mask) >> 56) < 6; )
|
|
m.magic = rng.sparse_rand<Bitboard>();
|
|
|
|
// A good magic must map every possible occupancy to an index that
|
|
// looks up the correct sliding attack in the attacks[s] database.
|
|
// Note that we build up the database for square 's' as a side
|
|
// effect of verifying the magic. Keep track of the attempt count
|
|
// and save it in epoch[], little speed-up trick to avoid resetting
|
|
// m.attacks[] after every failed attempt.
|
|
for (++cnt, i = 0; i < size; ++i)
|
|
{
|
|
unsigned idx = m.index(occupancy[i]);
|
|
|
|
if (epoch[idx] < cnt)
|
|
{
|
|
epoch[idx] = cnt;
|
|
m.attacks[idx] = reference[i];
|
|
}
|
|
else if (m.attacks[idx] != reference[i])
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|