1
0
Fork 0
mirror of https://github.com/sockspls/badfish synced 2025-05-03 18:19:35 +00:00
BadFish/src/movepick.cpp
nodchip 84f3e86790 Add NNUE evaluation
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
2020-08-06 16:37:45 +02:00

264 lines
8.7 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 "movepick.h"
namespace {
enum Stages {
MAIN_TT, CAPTURE_INIT, GOOD_CAPTURE, REFUTATION, QUIET_INIT, QUIET, BAD_CAPTURE,
EVASION_TT, EVASION_INIT, EVASION,
PROBCUT_TT, PROBCUT_INIT, PROBCUT,
QSEARCH_TT, QCAPTURE_INIT, QCAPTURE, QCHECK_INIT, QCHECK
};
// partial_insertion_sort() sorts moves in descending order up to and including
// a given limit. The order of moves smaller than the limit is left unspecified.
void partial_insertion_sort(ExtMove* begin, ExtMove* end, int limit) {
for (ExtMove *sortedEnd = begin, *p = begin + 1; p < end; ++p)
if (p->value >= limit)
{
ExtMove tmp = *p, *q;
*p = *++sortedEnd;
for (q = sortedEnd; q != begin && *(q - 1) < tmp; --q)
*q = *(q - 1);
*q = tmp;
}
}
} // namespace
/// Constructors of the MovePicker class. As arguments we pass information
/// to help it to return the (presumably) good moves first, to decide which
/// moves to return (in the quiescence search, for instance, we only want to
/// search captures, promotions, and some checks) and how important good move
/// ordering is at the current node.
/// MovePicker constructor for the main search
MovePicker::MovePicker(const Position& p, Move ttm, Depth d, const ButterflyHistory* mh, const LowPlyHistory* lp,
const CapturePieceToHistory* cph, const PieceToHistory** ch, Move cm, const Move* killers, int pl)
: pos(p), mainHistory(mh), lowPlyHistory(lp), captureHistory(cph), continuationHistory(ch),
ttMove(ttm), refutations{{killers[0], 0}, {killers[1], 0}, {cm, 0}}, depth(d), ply(pl) {
assert(d > 0);
stage = (pos.checkers() ? EVASION_TT : MAIN_TT) +
!(ttm && pos.pseudo_legal(ttm));
}
/// MovePicker constructor for quiescence search
MovePicker::MovePicker(const Position& p, Move ttm, Depth d, const ButterflyHistory* mh,
const CapturePieceToHistory* cph, const PieceToHistory** ch, Square rs)
: pos(p), mainHistory(mh), captureHistory(cph), continuationHistory(ch), ttMove(ttm), recaptureSquare(rs), depth(d) {
assert(d <= 0);
stage = (pos.checkers() ? EVASION_TT : QSEARCH_TT) +
!(ttm && (depth > DEPTH_QS_RECAPTURES || to_sq(ttm) == recaptureSquare)
&& pos.pseudo_legal(ttm));
}
/// MovePicker constructor for ProbCut: we generate captures with SEE greater
/// than or equal to the given threshold.
MovePicker::MovePicker(const Position& p, Move ttm, Value th, const CapturePieceToHistory* cph)
: pos(p), captureHistory(cph), ttMove(ttm), threshold(th) {
assert(!pos.checkers());
stage = PROBCUT_TT + !(ttm && pos.capture(ttm)
&& pos.pseudo_legal(ttm)
&& pos.see_ge(ttm, threshold));
}
/// MovePicker::score() assigns a numerical value to each move in a list, used
/// for sorting. Captures are ordered by Most Valuable Victim (MVV), preferring
/// captures with a good history. Quiets moves are ordered using the histories.
template<GenType Type>
void MovePicker::score() {
static_assert(Type == CAPTURES || Type == QUIETS || Type == EVASIONS, "Wrong type");
for (auto& m : *this)
if (Type == CAPTURES)
m.value = int(PieceValue[MG][pos.piece_on(to_sq(m))]) * 6
+ (*captureHistory)[pos.moved_piece(m)][to_sq(m)][type_of(pos.piece_on(to_sq(m)))];
else if (Type == QUIETS)
m.value = (*mainHistory)[pos.side_to_move()][from_to(m)]
+ 2 * (*continuationHistory[0])[pos.moved_piece(m)][to_sq(m)]
+ 2 * (*continuationHistory[1])[pos.moved_piece(m)][to_sq(m)]
+ 2 * (*continuationHistory[3])[pos.moved_piece(m)][to_sq(m)]
+ (*continuationHistory[5])[pos.moved_piece(m)][to_sq(m)]
+ (ply < MAX_LPH ? std::min(4, depth / 3) * (*lowPlyHistory)[ply][from_to(m)] : 0);
else // Type == EVASIONS
{
if (pos.capture(m))
m.value = PieceValue[MG][pos.piece_on(to_sq(m))]
- Value(type_of(pos.moved_piece(m)));
else
m.value = (*mainHistory)[pos.side_to_move()][from_to(m)]
+ (*continuationHistory[0])[pos.moved_piece(m)][to_sq(m)]
- (1 << 28);
}
}
/// MovePicker::select() returns the next move satisfying a predicate function.
/// It never returns the TT move.
template<MovePicker::PickType T, typename Pred>
Move MovePicker::select(Pred filter) {
while (cur < endMoves)
{
if (T == Best)
std::swap(*cur, *std::max_element(cur, endMoves));
if (*cur != ttMove && filter())
return *cur++;
cur++;
}
return MOVE_NONE;
}
/// MovePicker::next_move() is the most important method of the MovePicker class. It
/// returns a new pseudo legal move every time it is called until there are no more
/// moves left, picking the move with the highest score from a list of generated moves.
Move MovePicker::next_move(bool skipQuiets) {
top:
switch (stage) {
case MAIN_TT:
case EVASION_TT:
case QSEARCH_TT:
case PROBCUT_TT:
++stage;
return ttMove;
case CAPTURE_INIT:
case PROBCUT_INIT:
case QCAPTURE_INIT:
cur = endBadCaptures = moves;
endMoves = generate<CAPTURES>(pos, cur);
score<CAPTURES>();
++stage;
goto top;
case GOOD_CAPTURE:
if (select<Best>([&](){
return pos.see_ge(*cur, Value(-69 * cur->value / 1024)) ?
// Move losing capture to endBadCaptures to be tried later
true : (*endBadCaptures++ = *cur, false); }))
return *(cur - 1);
// Prepare the pointers to loop over the refutations array
cur = std::begin(refutations);
endMoves = std::end(refutations);
// If the countermove is the same as a killer, skip it
if ( refutations[0].move == refutations[2].move
|| refutations[1].move == refutations[2].move)
--endMoves;
++stage;
/* fallthrough */
case REFUTATION:
if (select<Next>([&](){ return *cur != MOVE_NONE
&& !pos.capture(*cur)
&& pos.pseudo_legal(*cur); }))
return *(cur - 1);
++stage;
/* fallthrough */
case QUIET_INIT:
if (!skipQuiets)
{
cur = endBadCaptures;
endMoves = generate<QUIETS>(pos, cur);
score<QUIETS>();
partial_insertion_sort(cur, endMoves, -3000 * depth);
}
++stage;
/* fallthrough */
case QUIET:
if ( !skipQuiets
&& select<Next>([&](){return *cur != refutations[0].move
&& *cur != refutations[1].move
&& *cur != refutations[2].move;}))
return *(cur - 1);
// Prepare the pointers to loop over the bad captures
cur = moves;
endMoves = endBadCaptures;
++stage;
/* fallthrough */
case BAD_CAPTURE:
return select<Next>([](){ return true; });
case EVASION_INIT:
cur = moves;
endMoves = generate<EVASIONS>(pos, cur);
score<EVASIONS>();
++stage;
/* fallthrough */
case EVASION:
return select<Best>([](){ return true; });
case PROBCUT:
return select<Best>([&](){ return pos.see_ge(*cur, threshold); });
case QCAPTURE:
if (select<Best>([&](){ return depth > DEPTH_QS_RECAPTURES
|| to_sq(*cur) == recaptureSquare; }))
return *(cur - 1);
// If we did not find any move and we do not try checks, we have finished
if (depth != DEPTH_QS_CHECKS)
return MOVE_NONE;
++stage;
/* fallthrough */
case QCHECK_INIT:
cur = moves;
endMoves = generate<QUIET_CHECKS>(pos, cur);
++stage;
/* fallthrough */
case QCHECK:
return select<Next>([](){ return true; });
}
assert(false);
return MOVE_NONE; // Silence warning
}