mirror of
https://github.com/HChaZZY/Stockfish.git
synced 2025-12-26 12:06:22 +08:00
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
171 lines
5.7 KiB
C++
171 lines
5.7 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 <cassert>
|
|
#include <vector>
|
|
#include <bitset>
|
|
|
|
#include "bitboard.h"
|
|
#include "types.h"
|
|
|
|
namespace {
|
|
|
|
// There are 24 possible pawn squares: files A to D and ranks from 2 to 7.
|
|
// Positions with the pawn on files E to H will be mirrored before probing.
|
|
constexpr unsigned MAX_INDEX = 2*24*64*64; // stm * psq * wksq * bksq = 196608
|
|
|
|
std::bitset<MAX_INDEX> KPKBitbase;
|
|
|
|
// A KPK bitbase index is an integer in [0, IndexMax] range
|
|
//
|
|
// Information is mapped in a way that minimizes the number of iterations:
|
|
//
|
|
// bit 0- 5: white king square (from SQ_A1 to SQ_H8)
|
|
// bit 6-11: black king square (from SQ_A1 to SQ_H8)
|
|
// bit 12: side to move (WHITE or BLACK)
|
|
// bit 13-14: white pawn file (from FILE_A to FILE_D)
|
|
// bit 15-17: white pawn RANK_7 - rank (from RANK_7 - RANK_7 to RANK_7 - RANK_2)
|
|
unsigned index(Color stm, Square bksq, Square wksq, Square psq) {
|
|
return int(wksq) | (bksq << 6) | (stm << 12) | (file_of(psq) << 13) | ((RANK_7 - rank_of(psq)) << 15);
|
|
}
|
|
|
|
enum Result {
|
|
INVALID = 0,
|
|
UNKNOWN = 1,
|
|
DRAW = 2,
|
|
WIN = 4
|
|
};
|
|
|
|
Result& operator|=(Result& r, Result v) { return r = Result(r | v); }
|
|
|
|
struct KPKPosition {
|
|
KPKPosition() = default;
|
|
explicit KPKPosition(unsigned idx);
|
|
operator Result() const { return result; }
|
|
Result classify(const std::vector<KPKPosition>& db);
|
|
|
|
Color stm;
|
|
Square ksq[COLOR_NB], psq;
|
|
Result result;
|
|
};
|
|
|
|
} // namespace
|
|
|
|
|
|
bool Bitbases::probe(Square wksq, Square wpsq, Square bksq, Color stm) {
|
|
|
|
assert(file_of(wpsq) <= FILE_D);
|
|
|
|
return KPKBitbase[index(stm, bksq, wksq, wpsq)];
|
|
}
|
|
|
|
|
|
void Bitbases::init() {
|
|
|
|
std::vector<KPKPosition> db(MAX_INDEX);
|
|
unsigned idx, repeat = 1;
|
|
|
|
// Initialize db with known win / draw positions
|
|
for (idx = 0; idx < MAX_INDEX; ++idx)
|
|
db[idx] = KPKPosition(idx);
|
|
|
|
// Iterate through the positions until none of the unknown positions can be
|
|
// changed to either wins or draws (15 cycles needed).
|
|
while (repeat)
|
|
for (repeat = idx = 0; idx < MAX_INDEX; ++idx)
|
|
repeat |= (db[idx] == UNKNOWN && db[idx].classify(db) != UNKNOWN);
|
|
|
|
// Fill the bitbase with the decisive results
|
|
for (idx = 0; idx < MAX_INDEX; ++idx)
|
|
if (db[idx] == WIN)
|
|
KPKBitbase.set(idx);
|
|
}
|
|
|
|
|
|
namespace {
|
|
|
|
KPKPosition::KPKPosition(unsigned idx) {
|
|
|
|
ksq[WHITE] = Square((idx >> 0) & 0x3F);
|
|
ksq[BLACK] = Square((idx >> 6) & 0x3F);
|
|
stm = Color ((idx >> 12) & 0x01);
|
|
psq = make_square(File((idx >> 13) & 0x3), Rank(RANK_7 - ((idx >> 15) & 0x7)));
|
|
|
|
// Invalid if two pieces are on the same square or if a king can be captured
|
|
if ( distance(ksq[WHITE], ksq[BLACK]) <= 1
|
|
|| ksq[WHITE] == psq
|
|
|| ksq[BLACK] == psq
|
|
|| (stm == WHITE && (pawn_attacks_bb(WHITE, psq) & ksq[BLACK])))
|
|
result = INVALID;
|
|
|
|
// Win if the pawn can be promoted without getting captured
|
|
else if ( stm == WHITE
|
|
&& rank_of(psq) == RANK_7
|
|
&& ksq[WHITE] != psq + NORTH
|
|
&& ( distance(ksq[BLACK], psq + NORTH) > 1
|
|
|| (distance(ksq[WHITE], psq + NORTH) == 1)))
|
|
result = WIN;
|
|
|
|
// Draw if it is stalemate or the black king can capture the pawn
|
|
else if ( stm == BLACK
|
|
&& ( !(attacks_bb<KING>(ksq[BLACK]) & ~(attacks_bb<KING>(ksq[WHITE]) | pawn_attacks_bb(WHITE, psq)))
|
|
|| (attacks_bb<KING>(ksq[BLACK]) & ~attacks_bb<KING>(ksq[WHITE]) & psq)))
|
|
result = DRAW;
|
|
|
|
// Position will be classified later
|
|
else
|
|
result = UNKNOWN;
|
|
}
|
|
|
|
Result KPKPosition::classify(const std::vector<KPKPosition>& db) {
|
|
|
|
// White to move: If one move leads to a position classified as WIN, the result
|
|
// of the current position is WIN. If all moves lead to positions classified
|
|
// as DRAW, the current position is classified as DRAW, otherwise the current
|
|
// position is classified as UNKNOWN.
|
|
//
|
|
// Black to move: If one move leads to a position classified as DRAW, the result
|
|
// of the current position is DRAW. If all moves lead to positions classified
|
|
// as WIN, the position is classified as WIN, otherwise the current position is
|
|
// classified as UNKNOWN.
|
|
const Result Good = (stm == WHITE ? WIN : DRAW);
|
|
const Result Bad = (stm == WHITE ? DRAW : WIN);
|
|
|
|
Result r = INVALID;
|
|
Bitboard b = attacks_bb<KING>(ksq[stm]);
|
|
|
|
while (b)
|
|
r |= stm == WHITE ? db[index(BLACK, ksq[BLACK] , pop_lsb(&b), psq)]
|
|
: db[index(WHITE, pop_lsb(&b), ksq[WHITE], psq)];
|
|
|
|
if (stm == WHITE)
|
|
{
|
|
if (rank_of(psq) < RANK_7) // Single push
|
|
r |= db[index(BLACK, ksq[BLACK], ksq[WHITE], psq + NORTH)];
|
|
|
|
if ( rank_of(psq) == RANK_2 // Double push
|
|
&& psq + NORTH != ksq[WHITE]
|
|
&& psq + NORTH != ksq[BLACK])
|
|
r |= db[index(BLACK, ksq[BLACK], ksq[WHITE], psq + NORTH + NORTH)];
|
|
}
|
|
|
|
return result = r & Good ? Good : r & UNKNOWN ? UNKNOWN : Bad;
|
|
}
|
|
|
|
} // namespace
|