Difference between revisions of "Stockfish NNUE"

From Chessprogramming wiki
Jump to: navigation, search
(20 intermediate revisions by 2 users not shown)
Line 5: Line 5:
 
'''Stockfish NNUE''',<br/>
 
'''Stockfish NNUE''',<br/>
 
a Stockfish branch by [[Hisayori Noda]] aka Nodchip, which uses [[NNUE|Efficiently Updatable Neural Networks]] - stylized as '''&#398;U&#1048;&#1048;''' or reversed as '''NNUE''' - to replace its standard [[Stockfish#Evaluation|evaluation]].
 
a Stockfish branch by [[Hisayori Noda]] aka Nodchip, which uses [[NNUE|Efficiently Updatable Neural Networks]] - stylized as '''&#398;U&#1048;&#1048;''' or reversed as '''NNUE''' - to replace its standard [[Stockfish#Evaluation|evaluation]].
NNUE, introduced in 2018 by [[Yu Nasu]] <ref>[[Yu Nasu]] ('''2018'''). ''&#398;U&#1048;&#1048; Efficiently Updatable Neural-Network based Evaluation Functions for Computer Shogi''.  Ziosoft Computer Shogi Club, [https://github.com/ynasu87/nnue/blob/master/docs/nnue.pdf pdf] (Japanese with English abstract)</ref>,  
+
NNUE, introduced in 2018 by [[Yu Nasu]] <ref>[[Yu Nasu]] ('''2018'''). ''&#398;U&#1048;&#1048; Efficiently Updatable Neural-Network based Evaluation Functions for Computer Shogi''.  Ziosoft Computer Shogi Club, [https://github.com/ynasu87/nnue/blob/master/docs/nnue.pdf pdf] (Japanese with English abstract) [https://github.com/asdfjkl/nnue GitHub - asdfjkl/nnue translation] </ref>,  
 
were previously successfully applied in [[Shogi]] evaluation functions embedded in a Stockfish based search <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=72754 The Stockfish of shogi] by [[Larry Kaufman]], [[CCC]], January 07, 2020</ref>, such as [[YaneuraOu]] <ref>[https://github.com/yaneurao/YaneuraOu GitHub - yaneurao/YaneuraOu: YaneuraOu is the World's Strongest Shogi engine(AI player), WCSC29 1st winner, educational and USI compliant engine]</ref>,
 
were previously successfully applied in [[Shogi]] evaluation functions embedded in a Stockfish based search <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=72754 The Stockfish of shogi] by [[Larry Kaufman]], [[CCC]], January 07, 2020</ref>, such as [[YaneuraOu]] <ref>[https://github.com/yaneurao/YaneuraOu GitHub - yaneurao/YaneuraOu: YaneuraOu is the World's Strongest Shogi engine(AI player), WCSC29 1st winner, educational and USI compliant engine]</ref>,
and [[Kristallweizen]] <ref>[https://github.com/Tama4649/Kristallweizen/ GitHub - Tama4649/Kristallweizen: 第29回世界コンピュータ将棋選手権 準優勝のKristallweizenです。]</ref>. He made an unbelievable prediction that NNUE can help to increase Stockfish strength by around 100 points, almost one year before revealing<ref>[http://yaneuraou.yaneu.com/2019/06/24/%E5%B0%86%E6%A3%8B%E3%82%BD%E3%83%95%E3%83%88%E9%96%8B%E7%99%BA%E8%80%85%E3%81%8Cstockfish%E3%81%AB%E8%B2%A2%E7%8C%AE%E3%81%99%E3%82%8B%E6%97%A5/ 将棋ソフト開発者がStockfishに貢献する日 The day when shogi software developers contribute to Stockfish] by [[Motohiro Isozaki]], June 2019</ref> <ref>[https://www.reddit.com/r/chess/comments/cltich/shogi_engine_developer_claims_he_can_make/ shogi engine developer claims he can make Stockfish stronger], [[Computer Chess Forums|Reddit]], August 2019</ref>.
+
and [[Kristallweizen]] <ref>[https://github.com/Tama4649/Kristallweizen/ GitHub - Tama4649/Kristallweizen: 第29回世界コンピュータ将棋選手権 準優勝のKristallweizenです。]</ref>. YaneuraOu's author [[Motohiro Isozaki]] made an unbelievable prediction that NNUE can help to increase Stockfish strength by around 100 points, almost one year before revealing <ref>[http://yaneuraou.yaneu.com/2019/06/24/%E5%B0%86%E6%A3%8B%E3%82%BD%E3%83%95%E3%83%88%E9%96%8B%E7%99%BA%E8%80%85%E3%81%8Cstockfish%E3%81%AB%E8%B2%A2%E7%8C%AE%E3%81%99%E3%82%8B%E6%97%A5/ 将棋ソフト開発者がStockfishに貢献する日 The day when shogi software developers contribute to Stockfish] by [[Motohiro Isozaki]], June 2019</ref> <ref>[https://www.reddit.com/r/chess/comments/cltich/shogi_engine_developer_claims_he_can_make/ shogi engine developer claims he can make Stockfish stronger], [[Computer Chess Forums|Reddit]], August 2019</ref>.
 
In 2019, Nodchip incorporated NNUE into Stockfish 10 - as a proof of concept, and with the intention to give something back to the Stockfish community <ref>[http://yaneuraou.yaneu.com/2020/06/19/stockfish-nnue-the-complete-guide/ Stockfish NNUE – The Complete Guide], June 19, 2020 (Japanese and English)</ref>.
 
In 2019, Nodchip incorporated NNUE into Stockfish 10 - as a proof of concept, and with the intention to give something back to the Stockfish community <ref>[http://yaneuraou.yaneu.com/2020/06/19/stockfish-nnue-the-complete-guide/ Stockfish NNUE – The Complete Guide], June 19, 2020 (Japanese and English)</ref>.
 
After support and announcements by [[Henk Drost]] in May 2020 <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74059 Stockfish NN release (NNUE)] by [[Henk Drost]], [[CCC]], May 31, 2020</ref>  
 
After support and announcements by [[Henk Drost]] in May 2020 <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74059 Stockfish NN release (NNUE)] by [[Henk Drost]], [[CCC]], May 31, 2020</ref>  
Line 16: Line 16:
  
 
=NNUE Structure=
 
=NNUE Structure=
The [[Neural Networks|neural network]] consists of four layers. The input layer is heavily overparametrized, feeding in the [[Board Representation|board representation]] for all king placements per side.
+
The [[Neural Networks|neural network]] consists of four layers. The input layer is heavily overparametrized, feeding in the [[Board Representation|board representation]] for all king placements per side
 +
<ref>[http://www.talkchess.com/forum3/viewtopic.php?f=7&t=74531&start=1 Re: NNUE accessible explanation] by [[Jonathan Rosenthal]], [[CCC]], July 23, 2020</ref>.
 +
 
 +
The so called '''HalfKP''' structure consists of two halves covering input layer and first hidden layer, each half of the input layer associated to one of the two [[King|kings]], cross coupled with the side to move or not to move halves of the first hidden layer.
 +
For each either black or white king placement, the 10 none king pieces on their particular squares are the boolean {0,1} inputs, along with a relict from Shogi piece drop (BONA_PIECE_ZERO),
 +
64 x (64 x 10 + 1) = 41,024 inputs for each half, which are multiplied by a 16-bit integer weight vector for 256 outputs per half, in total, 256 x 41,024 = 10,502,144 weights.
 +
As emphasized by [[Ronald de Man]] in a [[CCC]] forum discussion <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=7&t=75506&start=7 Re: NNUE Question - King Placements] by [[Ronald de Man|syzygy]], [[CCC]], October 23, 2020</ref>,
 +
the input weights are arranged in such a way, that [[Color Flipping|color flipped]] king-piece configurations in both halves share the same index.
 +
However, and that seems also a relict from Shogi with its [https://en.wikipedia.org/wiki/Rotational_symmetry 180 degrees rotational] 9x9 board symmetry, instead of [[Vertical Flipping|vertical flipping]] (xor 56), [[Flipping Mirroring and Rotating#Rotationby180degrees|rotation]] is applied (xor 63) <ref>[https://github.com/official-stockfish/Stockfish/issues/3021 NNUE eval rotate vs mirror · Issue #3021 · official-stockfish/Stockfish · GitHub] by [[Terje Kirstihagen]], August 17, 2020</ref>.
 +
 
 
The efficiency of [[NNUE]] is due to [[Incremental Updates|incremental update]] of the input layer outputs in [[Make Move|make]] and [[Unmake Move|unmake move]],
 
The efficiency of [[NNUE]] is due to [[Incremental Updates|incremental update]] of the input layer outputs in [[Make Move|make]] and [[Unmake Move|unmake move]],
where only a tiny fraction of its neurons need to be considered <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=7&t=74531&start=1 Re: NNUE accessible explanation] by [[Jonathan Rosenthal]], [[CCC]], July 23, 2020</ref>. The remaining three layers with 2x256x32, 32x32 and 32x1 weights are computational less expensive, hidden layer 1 and 2 with [https://en.wikipedia.org/wiki/Rectifier_(neural_networks) ReLu activation] <ref>[https://github.com/official-stockfish/Stockfish/blob/master/src/nnue/architectures/halfkp_256x2-32-32.h#L42 Stockfish/halfkp_256x2-32-32.h at master · official-stockfish/Stockfish · GitHub]</ref> <ref>[https://github.com/official-stockfish/Stockfish/blob/master/src/nnue/layers/clipped_relu.h#L82 Stockfish/clipped_relu.h at master · official-stockfish/Stockfish · GitHub]</ref>, best calculated using appropriate [[SIMD and SWAR Techniques|SIMD instructions]] performing fast [[Byte|8-bit]]/[[Word|16-bit]] integer vector arithmetic, like [[MMX]], [[SSE2]] or [[AVX2]] on [[x86]]/[[x86-64]], or, if available, [[AVX-512]].
+
where only a tiny fraction of its neurons need to be considered in case of none king moves.
 +
The remaining three layers with 2x256x32, 32x32 and 32x1 weights are computational less expensive, hidden layers apply a [https://en.wikipedia.org/wiki/Rectifier_(neural_networks) ReLu activation] <ref>[https://github.com/official-stockfish/Stockfish/blob/master/src/nnue/architectures/halfkp_256x2-32-32.h#L42 Stockfish/halfkp_256x2-32-32.h at master · official-stockfish/Stockfish · GitHub]</ref> <ref>[https://github.com/official-stockfish/Stockfish/blob/master/src/nnue/layers/clipped_relu.h#L82 Stockfish/clipped_relu.h at master · official-stockfish/Stockfish · GitHub]</ref>, best calculated using appropriate [[SIMD and SWAR Techniques|SIMD instructions]] performing fast [[Byte|8-bit]]/[[Word|16-bit]] integer vector arithmetic, like [[MMX]], [[SSE2]] or [[AVX2]] on [[x86]]/[[x86-64]], or, if available, [[AVX-512]].
  
 
[[FILE:StockfishNNUELayers.png|none|border|text-bottom|1024px]]  
 
[[FILE:StockfishNNUELayers.png|none|border|text-bottom|1024px]]  
NNUE layers in action <ref>Image courtesy Roman Zhukov, revised version of the image posted in [http://talkchess.com/forum3/viewtopic.php?f=2&t=74059&start=139 Re: Stockfish NN release (NNUE)] by Roman Zhukov, [[CCC]], June 17, 2020</ref>
+
NNUE layers in action <ref>Image courtesy Roman Zhukov, revised version of the image posted in [http://talkchess.com/forum3/viewtopic.php?f=2&t=74059&start=139 Re: Stockfish NN release (NNUE)] by Roman Zhukov, [[CCC]], June 17, 2020, labels corrected October 23, 2020, see [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=75506&start=1 Re: NNUE Question - King Placements] by [[Andrew Grant]], [[CCC]], October 23, 2020</ref>
 +
 
 +
Explanation by [[Ronald de Man]] <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=7&t=75506&start=9 Re: NNUE Question - King Placements] by [[Ronald de Man|syzygy]], [[CCC]], October 23, 2020</ref>, who did the Stockfish NNUE port to [[CFish]] <ref>[https://github.com/syzygy1/Cfish/blob/master/src/nnue.c Cfish/nnue.c at master · syzygy1/Cfish · GitHub]</ref>:
 +
The accumulator has a "white king" half and a "black king" half, where each half is a 256-element vector of 16-bit ints, which is equal to the sum of the weights of the "active" (pt, sq, ksq) features  plus a 256-element vector of 16-bit biases.
 +
 
 +
The "transform" step of the NNUE evaluation forms a 512-element vector of 8-bit ints where the first half is formed from the 256-element vector of the side to move and the second half is formed from the 256-element vector of the other side. In this step the 16-bit elements are clipped/clamped to a value from 0 to 127. This is the output of the input layer.
 +
 
 +
This 512-element vector of 8-bit ints is then multiplied by a 32x512 matrix of 8-bit weights to get a 32-element vector of 32-bit ints, to which a vector of 32-bit biases is added. The sum vector is divided by 64 and clipped/clamped to a 32-element vector of 8-bit ints from 0 to 127. This is the output of the first hidden layer.
 +
 
 +
The resulting 32-element vector of 8-bit ints is multiplied by a 32x32 matrix of 8-bit weights to get a 32-element vector of 32-bit ints, to which another vector of 32-bit biases is added. These ints are again divided by 64 and clipped/clamped to 32 8-bit ints from 0 to 127. This is the output of the second hidden layer.
  
[[Hisayori Noda|Nodchip]] explained on [https://en.wikipedia.org/wiki/Discord_(software) Discord] on June 2020 <ref>[http://talkchess.com/forum3/viewtopic.php?f=7&t=74780 Don't understand NNUE] by Lucasart, [[CCC]], August 14, 2020</ref>:
+
This 32-element vector of 8-bits ints is then multiplied by a 1x32 matrix of 8-bit weights (i.e. the inner product of two vectors is taken). This produces a 32-bit value to which a 32-bit bias is added. This gives the output of the output layer.
  
  41,024 = 64 * 641. 64 comes from the number of the cells where king may exist. 641 = 64 * 5 * 2 + 1. 64 here comes from the number of the cells where a piece other than king may exist. 5 is the number of piece types other than king. 2 is the number of the colors, white and black. 1 is a captured piece.
+
  The output of the output layer is divided by FV_SCALE = 16 to produce the NNUE evaluation. SF's evaluation then take some further steps such as adding a Tempo bonus (even though the NNUE evaluation inherently already takes into account the side to move in the "transform" step) and scaling the evaluation towards zero as rule50_count() approaches 50 moves.
"+ 1" is BONA_PIECE_ZERO.
 
Here "bona" means "[[Bonanza|bonanza]]" which is a popular computer [[Shogi|shogi]] engine. It introduced the feature "p" for the first time.
 
BonaPieces are contained in the evalList. It is updated by Position::do_move() and Position::undo_move(), and used by NNUE to calculate the network parameters between the input layer and the first hidden layer.
 
About the calculation, the following text will be helpful. This text is sent to RocketMiningPoo on Twitter.
 
"We add the i-th COLUMN of the W{0} to the z{0} for each i, where the i-th element is set to 1. And we subtract the i-th COULMN of the W{0} from the z{0} for each i, where the i-th element is set to 0. This operation is "accumulate" in your question." may be right... I hope that someone will double check.
 
  
 
=Network=
 
=Network=
Line 51: Line 65:
  
 
Being attracted by new advantages as well as being encouraged by some impressive successes, many developers joined or continued to work. The [[#Source|Official Stockfish]] repository shows the numbers of commits, ideas increased significantly after merging NNUE.
 
Being attracted by new advantages as well as being encouraged by some impressive successes, many developers joined or continued to work. The [[#Source|Official Stockfish]] repository shows the numbers of commits, ideas increased significantly after merging NNUE.
 
=Rotation vs Flip=
 
Since the 9x9 [[Shogi]] board has a centered king file and [[Castling|castling]] is not known in Shogi, [[Color Flipping|color flip]] versus[[Flipping Mirroring and Rotating#Rotationby180degrees|180 degree rotate]] differs in a [[Horizontal Mirroring|horizontal mirrored]] position from the other side's point of view, with otherwise identical playing options. The NNUE is trained and probed from the side to move point of view, where the used 180 degree rotation (xor 63 instead of 56) to flip sides looks rather strange for chess <ref>[https://github.com/official-stockfish/Stockfish/blob/615d98da2447e79ceceae205e0cd4e878115acc3/src/types.h#L323 Stockfish/types.h at 615d98da2447e79ceceae205e0cd4e878115acc3 · official-stockfish/Stockfish · GitHub]</ref>. i.e color flipping the black king from e8 to d1 rather than e1. Does it consider castling short to the queen side?
 
 
It is a little bit unclear, how that rotation rather than flip affects the playing strength <ref>[https://github.com/official-stockfish/Stockfish/issues/3021 NNUE eval rotate vs mirror · Issue #3021 · official-stockfish/Stockfish · GitHub] by [[Terje Kirstihagen]], August 17, 2020</ref> and whether NNUE for chess suffers from [https://en.wikipedia.org/wiki/Associative_visual_agnosia associative visual agnosia].
 
Maybe Fishtest needs to play many games with color-flipped openings, i.e. 1.e4 e5 and 1.e3 e5 2.e4, to look whether results differ or not.
 
Anyway, a fix from rotate to flip has to be done from producer and consumer sides, and is likely to void some training sessions.
 
  
 
=Suggestions=
 
=Suggestions=
Line 67: Line 74:
 
=See also=
 
=See also=
 
* [[:Category:NN|Category: Neural Network Engines]]
 
* [[:Category:NN|Category: Neural Network Engines]]
 +
* [[Fat Fritz#Fat Fritz 2|Fat Fritz 2.0]]
 
* [[Neural Networks]]
 
* [[Neural Networks]]
 
* [[NNUE]]
 
* [[NNUE]]
Line 73: Line 81:
 
=Forum Posts=  
 
=Forum Posts=  
 
==2020 ...==
 
==2020 ...==
===January ...===
+
===January ...===
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=72754 The Stockfish of shogi] by [[Larry Kaufman]], [[CCC]], January 07, 2020 » [[Shogi]]
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=72754 The Stockfish of shogi] by [[Larry Kaufman]], [[CCC]], January 07, 2020 » [[Shogi]]
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=74058 Stockfish NNUE] by [[Henk Drost]], [[CCC]], May 31, 2020 » [[Stockfish]]
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=74058 Stockfish NNUE] by [[Henk Drost]], [[CCC]], May 31, 2020 » [[Stockfish]]
Line 99: Line 107:
 
===August===
 
===August===
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74654 Repository for Stockfish+NNUE Android Builds] by [[Ted Summers|AdminX]], [[CCC]], August 02, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74654 Repository for Stockfish+NNUE Android Builds] by [[Ted Summers|AdminX]], [[CCC]], August 02, 2020
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74666 SF NNUE Problem] by Stephen Ham, [[CCC]], August 03, 2020
+
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74666 SF NNUE Problem] by [[Stephen Ham]], [[CCC]], August 03, 2020
 
* <span id="Continue"></span>[http://www.talkchess.com/forum3/viewtopic.php?f=7&t=74531&start=8 Re: NNUE accessible explanation] by [[Jonathan Rosenthal]], [[CCC]], August 03, 2020 » [[#NNUEaccExp|NNUE accessible explanation]]
 
* <span id="Continue"></span>[http://www.talkchess.com/forum3/viewtopic.php?f=7&t=74531&start=8 Re: NNUE accessible explanation] by [[Jonathan Rosenthal]], [[CCC]], August 03, 2020 » [[#NNUEaccExp|NNUE accessible explanation]]
 
* [https://groups.google.com/d/msg/fishcooking/6OI3AejYvpQ/dNmluMLBAgAJ <nowiki>[NNUE] Worker update on fishtest</nowiki>] by [[Joost VandeVondele]], [[Computer Chess Forums|FishCooking]], August 03, 2020
 
* [https://groups.google.com/d/msg/fishcooking/6OI3AejYvpQ/dNmluMLBAgAJ <nowiki>[NNUE] Worker update on fishtest</nowiki>] by [[Joost VandeVondele]], [[Computer Chess Forums|FishCooking]], August 03, 2020
Line 107: Line 115:
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74683 Is this SF NN almost like 20 MB book?] by [[Jouni Uski]], [[CCC]], August 04, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74683 Is this SF NN almost like 20 MB book?] by [[Jouni Uski]], [[CCC]], August 04, 2020
 
* [https://groups.google.com/d/msg/fishcooking/Kzw1W_Yr1d8/YNEmCqIyBAAJ NNUE evaluation merged in master] by [[Joost VandeVondele]], [[Computer Chess Forums|FishCooking]], August 06, 2020
 
* [https://groups.google.com/d/msg/fishcooking/Kzw1W_Yr1d8/YNEmCqIyBAAJ NNUE evaluation merged in master] by [[Joost VandeVondele]], [[Computer Chess Forums|FishCooking]], August 06, 2020
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74705 What happens with my hyperthreading?] by [[Kai Laskos]], [[CCC]], August 06, 2020 » [[Thread]]
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74722 Stockfish NNUE style] by Rowen, [[CCC]], August 08, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74722 Stockfish NNUE style] by Rowen, [[CCC]], August 08, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?t=74739 SF NNUE training questions] by [[Jouni Uski]], [[CCC]], August 10, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?t=74739 SF NNUE training questions] by [[Jouni Uski]], [[CCC]], August 10, 2020
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74765 Progress of Stockfish in 6 days] by [[Kai Laskos]], [[CCC]], August 12, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=74777 Neural Networks weights type] by [[Fabio Gobbato]], [[CCC]], August 13, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=74777 Neural Networks weights type] by [[Fabio Gobbato]], [[CCC]], August 13, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=74780 Don't understand NNUE] by Lucasart, [[CCC]], August 14, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=74780 Don't understand NNUE] by Lucasart, [[CCC]], August 14, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74901 SF+NNUE reach the ceiling?] by Corres, [[CCC]], August 27, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74901 SF+NNUE reach the ceiling?] by Corres, [[CCC]], August 27, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74933 The most stupid idea by the Stockfish Team] by Damir, [[CCC]], August 30, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74933 The most stupid idea by the Stockfish Team] by Damir, [[CCC]], August 30, 2020
===September ...===
+
===September===
 
* [https://groups.google.com/d/msg/fishcooking/TJHsiI61yQ4/liQoZ-AzAgAJ Stockfish 12] by [[Joost VandeVondele]], [[Computer Chess Forums|FishCooking]], September 02, 2020
 
* [https://groups.google.com/d/msg/fishcooking/TJHsiI61yQ4/liQoZ-AzAgAJ Stockfish 12] by [[Joost VandeVondele]], [[Computer Chess Forums|FishCooking]], September 02, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74974 Stockfish 12 is released today!] by Nay Lin Tun, [[CCC]], September 02, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74974 Stockfish 12 is released today!] by Nay Lin Tun, [[CCC]], September 02, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74978 Stockfish 12 has arrived!] by daniel71, [[CCC]], September 02, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74978 Stockfish 12 has arrived!] by daniel71, [[CCC]], September 02, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=75008 AVX2 optimized SF+NNUE and processor temperature]  by corres, [[CCC]], September 05, 2020 » [[AVX2]]
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=75008 AVX2 optimized SF+NNUE and processor temperature]  by corres, [[CCC]], September 05, 2020 » [[AVX2]]
===October ...===
+
===October===
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=75296 BONA_PIECE_ZERO] by [[Marco Belli|elcabesa]], [[CCC]], October 04, 2020
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=75296 BONA_PIECE_ZERO] by [[Marco Belli|elcabesa]], [[CCC]], October 04, 2020
 +
* [https://groups.google.com/d/msg/fishcooking/yjh1YOxy7nw/rJA6u1ODAAAJ SF NNUE/Classical] by [[Fauzi Akram Dabat|Fauzi]], [[Computer Chess Forums|FishCooking]], October 05, 2020
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=75415 How to scale stockfish NNUE score?] by [[Maksim Korzh]], [[CCC]], October 17, 2020 » [[Stockfish NNUE]], [[Scorpio#NNUE|Scorpio NNUE]]
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=75418 Embedding Stockfish NNUE to ANY CHESS ENGINE: YouTube series] by [[Maksim Korzh]], [[CCC]], October 17, 2020 » [[BBC#NNUE|BBC NNUE]]
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=75506 NNUE Question - King Placements] by [[Andrew Grant]], [[CCC]], October 23, 2020 » [[#NNUE Structure|NNUE Structure]]
 +
: [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=75506&start=7 Re: NNUE Question - King Placements] by [[Ronald de Man|syzygy]], [[CCC]], October 23, 2020
 +
: [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=75506&start=9 Re: NNUE Question - King Placements] by [[Ronald de Man|syzygy]], [[CCC]], October 23, 2020
 +
==2021 ...==
 +
===January===
 +
* [https://groups.google.com/g/fishcooking/c/cad1MGSdpU4/m/Ury4iBqSBgAJ Shouldn't positional attributes drive SF's NNUE input features (rather than king position)?] by [[Nick Pelling]], [[Computer Chess Forums|FishCooking]], January 10, 2021
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=76381 stockfish NNUE question] by [[Jon Dart]], [[CCC]], January 21, 2021
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=76437 256 in NNUE?] by Ted Wong, [[CCC]], January 28, 2021
 +
===February===
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=76537 Fat Fritz 2] by [[Jouni Uski]], [[CCC]], February 09, 2021 » [[Fat Fritz#Fat Fritz 2|Fat Fritz 2.0]]
  
 
=External Links=  
 
=External Links=  

Revision as of 21:30, 28 February 2021

Home * Engines * Stockfish * NNUE

Stockfish NNUE Logo [1]

Stockfish NNUE,
a Stockfish branch by Hisayori Noda aka Nodchip, which uses Efficiently Updatable Neural Networks - stylized as ƎUИИ or reversed as NNUE - to replace its standard evaluation. NNUE, introduced in 2018 by Yu Nasu [2], were previously successfully applied in Shogi evaluation functions embedded in a Stockfish based search [3], such as YaneuraOu [4], and Kristallweizen [5]. YaneuraOu's author Motohiro Isozaki made an unbelievable prediction that NNUE can help to increase Stockfish strength by around 100 points, almost one year before revealing [6] [7]. In 2019, Nodchip incorporated NNUE into Stockfish 10 - as a proof of concept, and with the intention to give something back to the Stockfish community [8]. After support and announcements by Henk Drost in May 2020 [9] and subsequent enhancements, Stockfish NNUE was established and recognized. In summer 2020, with more people involved in testing and training, the computer chess community bursts out enthusiastically due to its rapidly raising playing strength with different networks trained using a mixture of supervised and reinforcement learning methods. Despite the approximately halved search speed, Stockfish NNUE became stronger than its original [10].

In August 2020, Fishtest revealed Stockfish NNUE was stronger than the classical one at least 80 Elo [11]. In July 2020, the playing code of NNUE was put into the official Stockfish repository as a branch for further development and examination. In August that playing code merged to the master branch and become an official part of the engine. However, the training code still remained in Nodchip's repository [12] [13]. On September 02, 2020, Stockfish 12 was released with a huge jump in playing strength due to the introduction of NNUE and further tuning [14].

NNUE Structure

The neural network consists of four layers. The input layer is heavily overparametrized, feeding in the board representation for all king placements per side [15].

The so called HalfKP structure consists of two halves covering input layer and first hidden layer, each half of the input layer associated to one of the two kings, cross coupled with the side to move or not to move halves of the first hidden layer. For each either black or white king placement, the 10 none king pieces on their particular squares are the boolean {0,1} inputs, along with a relict from Shogi piece drop (BONA_PIECE_ZERO), 64 x (64 x 10 + 1) = 41,024 inputs for each half, which are multiplied by a 16-bit integer weight vector for 256 outputs per half, in total, 256 x 41,024 = 10,502,144 weights. As emphasized by Ronald de Man in a CCC forum discussion [16], the input weights are arranged in such a way, that color flipped king-piece configurations in both halves share the same index. However, and that seems also a relict from Shogi with its 180 degrees rotational 9x9 board symmetry, instead of vertical flipping (xor 56), rotation is applied (xor 63) [17].

The efficiency of NNUE is due to incremental update of the input layer outputs in make and unmake move, where only a tiny fraction of its neurons need to be considered in case of none king moves. The remaining three layers with 2x256x32, 32x32 and 32x1 weights are computational less expensive, hidden layers apply a ReLu activation [18] [19], best calculated using appropriate SIMD instructions performing fast 8-bit/16-bit integer vector arithmetic, like MMX, SSE2 or AVX2 on x86/x86-64, or, if available, AVX-512.

StockfishNNUELayers.png

NNUE layers in action [20]

Explanation by Ronald de Man [21], who did the Stockfish NNUE port to CFish [22]:

The accumulator has a "white king" half and a "black king" half, where each half is a 256-element vector of 16-bit ints, which is equal to the sum of the weights of the "active" (pt, sq, ksq) features  plus a 256-element vector of 16-bit biases.
The "transform" step of the NNUE evaluation forms a 512-element vector of 8-bit ints where the first half is formed from the 256-element vector of the side to move and the second half is formed from the 256-element vector of the other side. In this step the 16-bit elements are clipped/clamped to a value from 0 to 127. This is the output of the input layer.
This 512-element vector of 8-bit ints is then multiplied by a 32x512 matrix of 8-bit weights to get a 32-element vector of 32-bit ints, to which a vector of 32-bit biases is added. The sum vector is divided by 64 and clipped/clamped to a 32-element vector of 8-bit ints from 0 to 127. This is the output of the first hidden layer.
The resulting 32-element vector of 8-bit ints is multiplied by a 32x32 matrix of 8-bit weights to get a 32-element vector of 32-bit ints, to which another vector of 32-bit biases is added. These ints are again divided by 64 and clipped/clamped to 32 8-bit ints from 0 to 127. This is the output of the second hidden layer.
This 32-element vector of 8-bits ints is then multiplied by a 1x32 matrix of 8-bit weights (i.e. the inner product of two vectors is taken). This produces a 32-bit value to which a 32-bit bias is added. This gives the output of the output layer.
The output of the output layer is divided by FV_SCALE = 16 to produce the NNUE evaluation. SF's evaluation then take some further steps such as adding a Tempo bonus (even though the NNUE evaluation inherently already takes into account the side to move in the "transform" step) and scaling the evaluation towards zero as rule50_count() approaches 50 moves.

Network

All networks are built by some volunteers but not by any big community (differs from Leela Chess Zero) and can be uploaded into Fishtest for testing. Networks with good test results will be released officially on Fishtest website [23].

After long discussing the best way to publish networks with Stockfish [24], the developing team decided to embed the default network into Stockfish binaries, making sure NNUE always works as well as bringing more convenience to users.

Hybrid

In August 2020 a new patch changed Stockfish NNUE into a hybrid engine: it uses NNUE evaluation only on quite balanced material positions, otherwise uses the classical one. It could speed up to 10% and gain 20 Elo [25]. At that point, NNUE helped to increase already around 100 Elo for Stockfish. In the same month, Stockfish changed the default mode of using evaluation functions from classic to hybrid one, the last step to completely accept NNUE.

Strong Points

For Users

  • Runs with CPU only, and doesn't require expensive video cards, as well the need for installing drivers and 3rd specific libraries. Thus it is much easier to install (compared to engines using deep convolutional neural networks, such as Leela Chess Zero) and suitable for almost all modern computers. Using a GPU is even not practical for that small net - host-device-latency aka. kernel-launch-overhead [26] [27] to a then underemployed GPU are not sufficient for the intended NPS range [28]
  • Releases with only one network (via UCI options), that help to delete users' confusion from finding, selecting and setting up. The network is selected carefully from Fishtest

For Developers

  • Requires small training sets. Some high score networks can be built with the effort of one or a few people within a few days. It doesn't require the massive computing from a supercomputer and/or from community
  • Doesn’t require complicated systems such as a sophisticated client-server model to train networks. Just a single binary from Nodchip’ repo is enough to train
  • The NNUE code is independent and can be separated easily from the rest and integrated to other engines [29]

Being attracted by new advantages as well as being encouraged by some impressive successes, many developers joined or continued to work. The Official Stockfish repository shows the numbers of commits, ideas increased significantly after merging NNUE.

Suggestions

In reply to Unai Corzo, Motohiro Isozaki aka Yaneurao, suggested 3 techniques that applied successfully to Shogi and can be brought back to Stockfish NNUE and may improve it another 100 - 200 Elo [30] [31]:

See also

Forum Posts

2020 ...

January ...

July

Re: NNUE accessible explanation by Jonathan Rosenthal, CCC, July 23, 2020
Re: NNUE accessible explanation by Jonathan Rosenthal, CCC, July 24, 2020
Re: NNUE accessible explanation by Jonathan Rosenthal, CCC, August 03, 2020

August

Re: this will be the merge of a lifetime : SF 80 Elo+ by Henk Drost, CCC, August 04, 2020

September

October

Re: NNUE Question - King Placements by syzygy, CCC, October 23, 2020
Re: NNUE Question - King Placements by syzygy, CCC, October 23, 2020

2021 ...

January

February

External Links

Basics

Source

Networks

Rating

Misc

References

  1. Stockfish NNUE Logo from GitHub - nodchip/Stockfish: UCI chess engine by Nodchip
  2. Yu Nasu (2018). ƎUИИ Efficiently Updatable Neural-Network based Evaluation Functions for Computer Shogi. Ziosoft Computer Shogi Club, pdf (Japanese with English abstract) GitHub - asdfjkl/nnue translation
  3. The Stockfish of shogi by Larry Kaufman, CCC, January 07, 2020
  4. GitHub - yaneurao/YaneuraOu: YaneuraOu is the World's Strongest Shogi engine(AI player), WCSC29 1st winner, educational and USI compliant engine
  5. GitHub - Tama4649/Kristallweizen: 第29回世界コンピュータ将棋選手権 準優勝のKristallweizenです。
  6. 将棋ソフト開発者がStockfishに貢献する日 The day when shogi software developers contribute to Stockfish by Motohiro Isozaki, June 2019
  7. shogi engine developer claims he can make Stockfish stronger, Reddit, August 2019
  8. Stockfish NNUE – The Complete Guide, June 19, 2020 (Japanese and English)
  9. Stockfish NN release (NNUE) by Henk Drost, CCC, May 31, 2020
  10. Can the sardine! NNUE clobbers SF by Henk Drost, CCC, July 16, 2020
  11. Introducing NNUE Evaluation, August 06, 2020
  12. NNUE merge · Issue #2823 · official-stockfish/Stockfish · GitHub by Joost VandeVondele, July 25, 2020
  13. GitHub - nodchip/Stockfish: UCI chess engine by Nodchip
  14. Stockfish 12, The Stockfish Team, Stockfish Blog, September 02, 2020
  15. Re: NNUE accessible explanation by Jonathan Rosenthal, CCC, July 23, 2020
  16. Re: NNUE Question - King Placements by syzygy, CCC, October 23, 2020
  17. NNUE eval rotate vs mirror · Issue #3021 · official-stockfish/Stockfish · GitHub by Terje Kirstihagen, August 17, 2020
  18. Stockfish/halfkp_256x2-32-32.h at master · official-stockfish/Stockfish · GitHub
  19. Stockfish/clipped_relu.h at master · official-stockfish/Stockfish · GitHub
  20. Image courtesy Roman Zhukov, revised version of the image posted in Re: Stockfish NN release (NNUE) by Roman Zhukov, CCC, June 17, 2020, labels corrected October 23, 2020, see Re: NNUE Question - King Placements by Andrew Grant, CCC, October 23, 2020
  21. Re: NNUE Question - King Placements by syzygy, CCC, October 23, 2020
  22. Cfish/nnue.c at master · syzygy1/Cfish · GitHub
  23. Neural Net download and statistics
  24. Improve dealing with the default net? Issue ##3030 · official-stockfish/Stockfish · GitHub by Joost VandeVondele, August 19, 2020
  25. NNUE evaluation threshold by MJZ1977 · Pull Request #2916 · official-stockfish/Stockfish · GitHub, August 06, 2020
  26. AB search with NN on GPU... by Srdja Matovic, CCC, August 13, 2020 » GPU
  27. kernel launch latency - CUDA / CUDA Programming and Performance - NVIDIA Developer Forums by LukeCuda, June 18, 2018
  28. stockfish with graphics card by h1a8, CCC, August 06, 2020
  29. NNUE Engines
  30. 3 technologies in shogi AI that could be used for chess AI, Motohiro Isozaki, August 2020
  31. GitHub - NNUE ideas and discussion (post-merge). #2915, August 2020
  32. An info by Sylwy, CCC, July 25, 2020
  33. You can now look inside NNUE and look at its Per square value estimation by Henk Drost, CCC, August 04, 2020

Up one Level