Difference between revisions of "Leela Chess Zero"

From Chessprogramming wiki
Jump to: navigation, search
(41 intermediate revisions by the same user not shown)
Line 14: Line 14:
 
Leela Chess Zero consists of an executable to play or analyze [[Chess Game|games]], initially dubbed '''LCZero''', soon rewritten by a team around [[Alexander Lyashuk]] for better performance and then called '''Lc0''' <ref>[https://github.com/LeelaChessZero/lc0/wiki/lc0-transition lc0 transition · LeelaChessZero/lc0 Wiki · GitHub]</ref> <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=68094&start=91 Re: TCEC season 13, 2 NN engines will be participating, Leela and Deus X] by [[Gian-Carlo Pascutto]], [[CCC]], August 03, 2018</ref>. This executable, the actual chess engine, performs the [[Monte-Carlo Tree Search|MCTS]] and reads the self-taught [[Neural Networks#Convolutional|CNN]], which weights are persistent in a separate file.
 
Leela Chess Zero consists of an executable to play or analyze [[Chess Game|games]], initially dubbed '''LCZero''', soon rewritten by a team around [[Alexander Lyashuk]] for better performance and then called '''Lc0''' <ref>[https://github.com/LeelaChessZero/lc0/wiki/lc0-transition lc0 transition · LeelaChessZero/lc0 Wiki · GitHub]</ref> <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=68094&start=91 Re: TCEC season 13, 2 NN engines will be participating, Leela and Deus X] by [[Gian-Carlo Pascutto]], [[CCC]], August 03, 2018</ref>. This executable, the actual chess engine, performs the [[Monte-Carlo Tree Search|MCTS]] and reads the self-taught [[Neural Networks#Convolutional|CNN]], which weights are persistent in a separate file.
 
Lc0 is written in [[Cpp#14|C++14]] and may be compiled for various platforms and backends. Since deep CNN approaches are best suited to run massively in parallel on [[GPU|GPUs]] to perform all the [[Float|floating point]] [https://en.wikipedia.org/wiki/Dot_product dot products] for thousands of neurons,  
 
Lc0 is written in [[Cpp#14|C++14]] and may be compiled for various platforms and backends. Since deep CNN approaches are best suited to run massively in parallel on [[GPU|GPUs]] to perform all the [[Float|floating point]] [https://en.wikipedia.org/wiki/Dot_product dot products] for thousands of neurons,  
the preferred target platforms are [[Nvidia]] [[GPU|GPUs]] supporting [https://en.wikipedia.org/wiki/CUDA CUDA] and cuDNN libraries <ref>[https://developer.nvidia.com/cudnn NVIDIA cuDNN | NVIDIA Developer]</ref>.
+
the preferred target platforms are [[Nvidia]] [[GPU|GPUs]] supporting [https://en.wikipedia.org/wiki/CUDA CUDA] and [https://en.wikipedia.org/wiki/CuDNN CuDNN] libraries <ref>[https://developer.nvidia.com/cudnn NVIDIA cuDNN | NVIDIA Developer]</ref>. [[Ankan Banerjee]] wrote the CuDNN backend code, also shared by [[Deus X]] and [[Allie]] <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=71822&start=48 Re: My failed attempt to change TCEC NN clone rules] by [[Adam Treat]], [[CCC]], September 19, 2019</ref>. None CUDA compliant GPUs ([[AMD]]) are supported through [[OpenCL]], while much slower pure CPU binaries are possible using [https://en.wikipedia.org/wiki/Basic_Linear_Algebra_Subprograms BLAS], target systems with or without a [https://en.wikipedia.org/wiki/Video_card graphics card] (GPU) are [[Linux]], [[Mac OS]] and [[Windows]] computers, or BLAS only the [[Raspberry Pi]].
None CUDA compliant GPUs ([[AMD]]) are supported through [https://en.wikipedia.org/wiki/OpenCL OpenCL], while much slower pure CPU binaries are possible using [https://en.wikipedia.org/wiki/Basic_Linear_Algebra_Subprograms BLAS], target systems with or without a [https://en.wikipedia.org/wiki/Video_card graphics card] (GPU) are [[Linux]], [[Mac OS]] and [[Windows]] computers, or BLAS only the [[Raspberry Pi]].
 
  
 
=Description=
 
=Description=
Line 27: Line 26:
  
 
==Network==
 
==Network==
While [[AlphaGo]] used two disjoint networks for policy and value, [[AlphaZero]] as well as Leela Chess Zero, share a common "body" connected to disjoint policy and value "heads". The “body” consists of spatial 8x8 input planes, followed by convolutional layers with B [[Neural Networks#Residual|residual]] blocks times 3x3xF filters. FxB specifies the model and size of the CNN (64x6, 128x10, 192x15, 256x20 were used).  
+
While [[AlphaGo]] used two disjoint networks for policy and value, [[AlphaZero]] as well as Leela Chess Zero, share a common "body" connected to disjoint policy and value "heads". The “body” consists of spatial 8x8 planes, using B [[Neural Networks#Residual|residual]] blocks with F filters of kernel size 3x3, stride 1. So far, model sizes FxB of 64x6, 128x10, 192x15, and 256x20 were used.  
Concerning [[Nodes per Second|nodes per second]] of the MCTS, smaller models are faster to calculate than larger models. They are faster to train and one may earlier recognize progress, but they will also saturate earlier so that at some point more training will no longer improve the engine.  
+
 
Larger and deeper network models will improve the receptivity, the amount of knowledge and pattern to extract from the training samples, with potential for a [[Playing Strength|stronger]] engine.  
+
Concerning [[Nodes per Second|nodes per second]] of the MCTS, smaller models are faster to calculate than larger models. They are faster to train and will earlier recognize progress, but they will also saturate earlier so that at some point more training will no longer improve the engine. Larger and deeper network models will improve the receptivity, the amount of knowledge and pattern to extract from the training samples, with potential for a [[Playing Strength|stronger]] engine.  
As a further improvement, Leele Chess Zero applies the ''Squeeze and Excite'' (SE) extension to the residual block architecture <ref>[https://github.com/LeelaChessZero/lc0/wiki/Technical-Explanation-of-Leela-Chess-Zero Technical Explanation of Leela Chess Zero · LeelaChessZero/lc0 Wiki · GitHub]</ref> <ref>[https://towardsdatascience.com/squeeze-and-excitation-networks-9ef5e71eacd7 Squeeze-and-Excitation Networks – Towards Data Science] by [http://plpp.de/ Paul-Louis Pröve], October 17, 2017</ref>.
+
As a further improvement, Leele Chess Zero applies the ''Squeeze and Excite'' (SE) extension to the residual block architecture <ref>[https://github.com/LeelaChessZero/lc0/wiki/Technical-Explanation-of-Leela-Chess-Zero Technical Explanation of Leela Chess Zero · LeelaChessZero/lc0 Wiki · GitHub]</ref> <ref>[https://towardsdatascience.com/squeeze-and-excitation-networks-9ef5e71eacd7 Squeeze-and-Excitation Networks – Towards Data Science] by [http://plpp.de/ Paul-Louis Pröve], October 17, 2017</ref>. The body is connected to both the policy "head" for the move probability distribution, and the value "head" for the evaluation score aka [[Pawn Advantage, Win Percentage, and Elo|winning probability]] of the current position and up to seven predecessor positions on the input planes.
The body is fully connected to both the policy "head" for the move probability distribution, and value "head" for the evaluation score aka [[Pawn Advantage, Win Percentage, and Elo|winning probability]] of the the the current position and up to seven predecessor positions on the input planes.
 
  
 
==Training==
 
==Training==
Like in [[AlphaZero]], the '''Zero''' suffix implies no other initial knowledge than the rules of the game, to build a superhuman player, starting with truly random self-play games to apply [[Reinforcement Learning|reinforcement learning]] based on the outcome of that games.
+
Like in [[AlphaZero]], the '''Zero''' suffix implies no other initial knowledge than the rules of the game, to build a superhuman player, starting with truly random self-play games to apply [[Reinforcement Learning|reinforcement learning]] based on the outcome of that games. However, there are derived approaches, such as [[Albert Silver|Albert Silver's]] [[Deus X]], trying to take a short-cut by initially using [[Supervised Learning|supervised learning]] techniques, such as feeding in high quality games played by other strong chess playing entities, or huge records of positions with a given preferred move.
However, there are derived approaches, such as [[Albert Silver|Albert Silver's]] [[Deus X]], trying to take a short-cut by initially using [[Supervised Learning|supervised learning]] techniques, such as feeding in high quality games played by other strong chess playing entities, or huge records of positions with a given preferred move.
 
 
The unsupervised training of the NN is about to minimize the [https://en.wikipedia.org/wiki/Norm_(mathematics)#Euclidean_norm L2-norm] of the [https://en.wikipedia.org/wiki/Mean_squared_error mean squared error] loss of the value output and the policy loss. Further there are experiments to train the value head against not the game outcome, but against the accumulated value for a position after exploring some number of nodes with [[UCT]] <ref>[https://medium.com/oracledevs/lessons-from-alphazero-part-4-improving-the-training-target-6efba2e71628 Lessons From AlphaZero (part 4): Improving the Training Target] by [https://blogs.oracle.com/author/vish-abrams Vish Abrams], [https://blogs.oracle.com/ Oracle Blog], June 27, 2018</ref>.
 
The unsupervised training of the NN is about to minimize the [https://en.wikipedia.org/wiki/Norm_(mathematics)#Euclidean_norm L2-norm] of the [https://en.wikipedia.org/wiki/Mean_squared_error mean squared error] loss of the value output and the policy loss. Further there are experiments to train the value head against not the game outcome, but against the accumulated value for a position after exploring some number of nodes with [[UCT]] <ref>[https://medium.com/oracledevs/lessons-from-alphazero-part-4-improving-the-training-target-6efba2e71628 Lessons From AlphaZero (part 4): Improving the Training Target] by [https://blogs.oracle.com/author/vish-abrams Vish Abrams], [https://blogs.oracle.com/ Oracle Blog], June 27, 2018</ref>.
  
The distributed training is realized with an sophisticated [https://en.wikipedia.org/wiki/Client%E2%80%93server_model client-server model].  
+
The distributed training is realized with an sophisticated [https://en.wikipedia.org/wiki/Client%E2%80%93server_model client-server model]. The client, written entirely in the [[Go (Programming Language)|Go programming language]], incorporates Lc0 to produce self-play games.Controlled by the server, the client may download the latest network, will start self-playing, and uploading games to the server, who on the other hand will regularly produce and distribute new neural network weights after a certain amount of games available from contributors. The training software consists of [[Python]] code, the pipeline requires [https://en.wikipedia.org/wiki/NumPy NumPy] and [https://en.wikipedia.org/wiki/TensorFlow TensorFlow] running on [[Linux]] <ref>[https://github.com/LeelaChessZero/lczero-training GitHub - LeelaChessZero/lczero-training: For code etc relating to the network training process.]</ref>.  
The client, written entirely in the [[Go (Programming Language)|Go programming language]], incorporates Lc0 to produce self-play games.
 
Controlled by the server, the client may download the latest network, will start self-playing, and uploading games to the server, who on the other hand will regularly produce and distribute new neural network weights after a certain amount of games available from contributors.  
 
The training software consists of [[Python]] code, the pipeline requires [https://en.wikipedia.org/wiki/NumPy NumPy] and [https://en.wikipedia.org/wiki/TensorFlow TensorFlow] running on [[Linux]] <ref>[https://github.com/LeelaChessZero/lczero-training GitHub - LeelaChessZero/lczero-training: For code etc relating to the network training process.]</ref>.  
 
 
The server is written in [[Go (Programming Language)|Go]] along with [[Python]] and [https://en.wikipedia.org/wiki/Shell_script shell scripts].
 
The server is written in [[Go (Programming Language)|Go]] along with [[Python]] and [https://en.wikipedia.org/wiki/Shell_script shell scripts].
  
 +
=Structure Diagrams=
 +
[[FILE:lc0diagram.png|none|border|text-bottom]]
 +
Related to [[TCEC]] clone discussions concerning [[Deus X]] and [[Allie]] aka [[Allie#AllieStein|AllieStein]],<br/>
 +
[[Alexander Lyashuk]] published diagrams with all components of the affected engines,<br/>
 +
The above shows the common legend, and the structure of all Leela Chess Zero's components based on current Lc0 engine <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=71822 My failed attempt to change TCEC NN clone rules] by [[Alexander Lyashuk]], [[CCC]], September 14, 2019 » [[TCEC]]</ref>
 +
[[FILE:Lczero.png|none|border|text-bottom|670px]]
 +
Same diagram, but initial LCZero engine, which played [[TCEC Season 12]] <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=71822 My failed attempt to change TCEC NN clone rules] by [[Alexander Lyashuk]], [[CCC]], September 14, 2019 » [[TCEC]]</ref>
 
=See also=
 
=See also=
 +
* [[Allie]]
 
* [[AlphaZero]]
 
* [[AlphaZero]]
 +
* [[Fat Fritz]]
 +
* [[Deus X]]
 
* [[Leela Zero]]
 
* [[Leela Zero]]
 
* [[Leila]]
 
* [[Leila]]
Line 63: Line 67:
 
* [http://www.talkchess.com/forum/viewtopic.php?t=66945 LCZero: Progress and Scaling. Relation to CCRL Elo] by [[Kai Laskos]], [[CCC]], March 28, 2018 » [[Playing Strength]]
 
* [http://www.talkchess.com/forum/viewtopic.php?t=66945 LCZero: Progress and Scaling. Relation to CCRL Elo] by [[Kai Laskos]], [[CCC]], March 28, 2018 » [[Playing Strength]]
 
* [http://www.talkchess.com/forum/viewtopic.php?t=67013 What does LCzero learn?] by [[Uri Blass]], [[CCC]], April 05, 2018
 
* [http://www.talkchess.com/forum/viewtopic.php?t=67013 What does LCzero learn?] by [[Uri Blass]], [[CCC]], April 05, 2018
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=67044 How to play vs LCZero with Cute Chess gui] by Hai, [[CCC]], April 08, 2018 » [[Cute Chess]]
 
* [http://www.talkchess.com/forum/viewtopic.php?t=67075 LCZero in Aquarium / Fritz] by [[Carl Bicknell]], [[CCC]], April 11, 2018
 
* [http://www.talkchess.com/forum/viewtopic.php?t=67075 LCZero in Aquarium / Fritz] by [[Carl Bicknell]], [[CCC]], April 11, 2018
 
* [http://www.talkchess.com/forum/viewtopic.php?t=67087 LCZero on 10x128 now] by [[Gary Linscott|Gary]], [[CCC]], April 12, 2018  
 
* [http://www.talkchess.com/forum/viewtopic.php?t=67087 LCZero on 10x128 now] by [[Gary Linscott|Gary]], [[CCC]], April 12, 2018  
 
* [http://www.talkchess.com/forum/viewtopic.php?t=67092 lczero faq] by Duncan Roberts, [[CCC]], April 13, 2018
 
* [http://www.talkchess.com/forum/viewtopic.php?t=67092 lczero faq] by Duncan Roberts, [[CCC]], April 13, 2018
 +
* [http://www.talkchess.com/forum/viewtopic.php?t=67104 Run LC Zero in LittleBlitzerGUI] by [[Stefan Pohl]], [[CCC]], April 14, 2018 » [[LittleBlitzer]]
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=67121 LC0 - how to catch up?] by [[Srdja Matovic]], [[CCC]], April 16, 2018
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=67300 Leela on more then one GPU?] by [[Karlo Balla]], [[CCC]], May 01, 2018
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=67646 New CLOP settings give Leela huge tactics boost] by [[Albert Silver]], [[CCC]], June 04, 2018 » [[CLOP]]
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=67668 First Win by Leela Chess Zero against Stockfish dev] by [[Ankan Banerjee]], [[CCC]], June 07, 2018 » [[Stockfish]]
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=67718 what may be two firsts...] by [[Michael Byrne|Michael B]], [[CCC]], June 13, 2018 » [[DGT Pi]]
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=67728 LcZero and STS] by [[Ed Schroder|Ed Schröder]], [[CCC]], June 14, 2018 » [[Strategic Test Suite]]
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=67728 LcZero and STS] by [[Ed Schroder|Ed Schröder]], [[CCC]], June 14, 2018 » [[Strategic Test Suite]]
 
* [https://groups.google.com/d/msg/lczero/S-rhiPLnbHg/XY9-Z1LWCAAJ Who entered Leela into WCCC? Bad idea!!] by [[Chris Whittington]], [[Computer Chess Forums|LCZero Forum]], June  23, 2018 » [[WCCC 2018]]
 
* [https://groups.google.com/d/msg/lczero/S-rhiPLnbHg/XY9-Z1LWCAAJ Who entered Leela into WCCC? Bad idea!!] by [[Chris Whittington]], [[Computer Chess Forums|LCZero Forum]], June  23, 2018 » [[WCCC 2018]]
Line 76: Line 87:
 
* [https://groups.google.com/d/msg/lczero/vGdNYW-Ou58/Kh0GCj2OCgAJ Has Silver written any code for "his" ZeusX?] by [[Chris Whittington]], [[Computer Chess Forums|LCZero Forum]], July 31, 2018  
 
* [https://groups.google.com/d/msg/lczero/vGdNYW-Ou58/Kh0GCj2OCgAJ Has Silver written any code for "his" ZeusX?] by [[Chris Whittington]], [[Computer Chess Forums|LCZero Forum]], July 31, 2018  
 
: [https://groups.google.com/d/msg/lczero/vGdNYW-Ou58/-icwb0pjDAAJ Re: Has Silver written any code for "his" ZeusX?] by [[Alexander Lyashuk]], [[Computer Chess Forums|LCZero Forum]], August 02, 2018  
 
: [https://groups.google.com/d/msg/lczero/vGdNYW-Ou58/-icwb0pjDAAJ Re: Has Silver written any code for "his" ZeusX?] by [[Alexander Lyashuk]], [[Computer Chess Forums|LCZero Forum]], August 02, 2018  
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=68448 How good is the RTX 2080 Ti for Leela?] by Hai, September 15, 2018 <ref>[https://en.wikipedia.org/wiki/GeForce_20_series GeForce 20 series from Wikipedia]</ref>
 +
: [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=68448&start=2 Re: How good is the RTX 2080 Ti for Leela?] by [[Ankan Banerjee]], [[CCC]], September 16, 2018
 +
: [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=68448&start=9 Re: How good is the RTX 2080 Ti for Leela?] by [[Ankan Banerjee]], [[CCC]], September 17, 2018 <ref>[https://en.wikipedia.org/wiki/Multiply%E2%80%93accumulate_operation Multiply–accumulate operation - Wikipedia]</ref>
 +
: [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=68448&start=37 Re: How good is the RTX 2080 Ti for Leela?] by [[Ankan Banerjee]], [[CCC]], October 28, 2018
 +
: [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=68448&start=44 Re: How good is the RTX 2080 Ti for Leela?] by [[Ankan Banerjee]], [[CCC]], November 15, 2018
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=68511 LC0 0.18rc1 released] by [[Günther Simon]], [[CCC]], September 25, 2018
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=68511 LC0 0.18rc1 released] by [[Günther Simon]], [[CCC]], September 25, 2018
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=68973 My non-OC RTX 2070 is very fast with Lc0] by [[Kai Laskos]], [[CCC]], November 19, 2018 <ref>[https://www.nvidia.com/en-us/geforce/graphics-cards/rtx-2070/ GeForce RTX 2070 Graphics Card | NVIDIA]</ref>
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=68973 My non-OC RTX 2070 is very fast with Lc0] by [[Kai Laskos]], [[CCC]], November 19, 2018 <ref>[https://www.nvidia.com/en-us/geforce/graphics-cards/rtx-2070/ GeForce RTX 2070 Graphics Card | NVIDIA]</ref>
Line 86: Line 102:
 
==2019==
 
==2019==
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=69478 LCZero FAQ is missing one important fact] by [[Jouni Uski]], [[CCC]], January 01, 2019 » [[GPU]]
 
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=69478 LCZero FAQ is missing one important fact] by [[Jouni Uski]], [[CCC]], January 01, 2019 » [[GPU]]
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=69552 Leela on a weak pc, question] by chessico, [[CCC]], January 09, 2019
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=69622 Can somebody explain what makes Leela as strong as Komodo?] by Chessqueen, [[CCC]], January 16, 2019
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=69668 A0 policy head ambiguity] by [[Daniel Shawul]], [[CCC]], January 21, 2019 » [[AlphaZero]]
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=69672 Schizophrenic rating model for Leela] by [[Kai Laskos]], [[CCC]], January 21, 2019 » [[Match Statistics]]
 +
* [http://forum.computerschach.de/cgi-bin/mwf/topic_show.pl?tid=10194 Leela Zero (Lc0) - NVIDIA Geforce RTX 2060] by [[Andreas Strangmüller]], [[Computer Chess Forums|CSS Forum]], January 29, 2019
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=69820 11258-32x4-se distilled network released] by [[Dietrich Kappe]], [[CCC]], February 03, 2019
 +
* [http://forum.computerschach.de/cgi-bin/mwf/topic_show.pl?tid=10213 Lc0 setup Hilfe] by [[Clemens Keck]], [[Computer Chess Forums|CSS Forum]], February 07, 2019 (German)
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=69852 Lc0 - macOS binary requested] by Steppenwolf, [[CCC]], February 09, 2019 » [[Mac OS]]
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=69957 Thanks for LC0] by [[Peter Berger]], [[CCC]], February 19, 2019
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=70069&start=10 Re: Training the trainer: how is it done for Stockfish?] by Graham Jones, [[CCC]], March 03, 2019 » [[Monte-Carlo Tree Search]], [[Stockfish]]
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=70350 Lc0 51010] by [[Larry Kaufman]], [[CCC]], March 29, 2019
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=70451 32930 Boost network available] by [[Dietrich Kappe]], [[CCC]], April 09, 2019
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=71209 Lc0 question] by [[Larry Kaufman]], [[CCC]], July 06, 2019
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=71651 Some newbie questions about lc0] by [[Pham Hong Nguyen|Nguyen Pham]], [[CCC]], August 25, 2019
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=71686 Lc0 Evaluation Explanation] by Hamster, [[CCC]], August 29, 2019
 +
: [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=71686&start=14 Re: Lc0 Evaluation Explanation] by [[Alexander Lyashuk]], [[CCC]], September 03, 2019
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=71822 My failed attempt to change TCEC NN clone rules] by [[Alexander Lyashuk]], [[CCC]], September 14, 2019 » [[TCEC]]
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=72640 Best Nets for Lc0 Page] by [[Ted Summers]], [[CCC]], December 23, 2019
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=2&t=72685 Correct LC0 syntax for multiple GPUs] by [[Dann Corbit]], [[CCC]], December 30, 2019
  
 
=Blog Posts=
 
=Blog Posts=
 +
==2018==
 
* [https://medium.com/oracledevs/lessons-from-implementing-alphazero-7e36e9054191 Lessons From Implementing AlphaZero] by [https://medium.com/@akprasad Aditya Prasad], [https://blogs.oracle.com/ Oracle Blog], June 05, 2018
 
* [https://medium.com/oracledevs/lessons-from-implementing-alphazero-7e36e9054191 Lessons From Implementing AlphaZero] by [https://medium.com/@akprasad Aditya Prasad], [https://blogs.oracle.com/ Oracle Blog], June 05, 2018
 
: [https://medium.com/oracledevs/lessons-from-alphazero-connect-four-e4a0ae82af68 Lessons from AlphaZero: Connect Four] by [https://medium.com/@akprasad Aditya Prasad], [https://blogs.oracle.com/ Oracle Blog], June 13, 2018
 
: [https://medium.com/oracledevs/lessons-from-alphazero-connect-four-e4a0ae82af68 Lessons from AlphaZero: Connect Four] by [https://medium.com/@akprasad Aditya Prasad], [https://blogs.oracle.com/ Oracle Blog], June 13, 2018
Line 101: Line 137:
 
* [http://blog.lczero.org/2018/11/lc0-v0190-has-been-released.html c0 v0.19.0 has been released] by [[Alexander Lyashuk|crem]], [[Leela Chess Zero|LCZero blog]], November 9, 2018
 
* [http://blog.lczero.org/2018/11/lc0-v0190-has-been-released.html c0 v0.19.0 has been released] by [[Alexander Lyashuk|crem]], [[Leela Chess Zero|LCZero blog]], November 9, 2018
 
* [http://blog.lczero.org/2018/12/alphazero-paper-and-lc0-v0191.html AlphaZero paper, and Lc0 v0.19.1] by [[Alexander Lyashuk|crem]], [[Leela Chess Zero|LCZero blog]], December 07, 2018
 
* [http://blog.lczero.org/2018/12/alphazero-paper-and-lc0-v0191.html AlphaZero paper, and Lc0 v0.19.1] by [[Alexander Lyashuk|crem]], [[Leela Chess Zero|LCZero blog]], December 07, 2018
 +
==2019==
 +
* [https://blog.lczero.org/2019/05/lc0-won-tcec-15.html Lc0 won TCEC 15] by [[Alexander Lyashuk|crem]], [[Leela Chess Zero|LCZero blog]], May 28, 2019 » [[TCEC Season 15]]
 +
* [https://blog.lczero.org/2019/07/end-of-era.html End of era] by [[Alexander Lyashuk|crem]], [[Leela Chess Zero|LCZero blog]], July 26, 2019
 +
* [https://blog.lczero.org/2019/12/lc0-v0230-has-been-released.html Lc0 v0.23.0 has been released] by [[Alexander Lyashuk|crem]], [[Leela Chess Zero|LCZero blog]], December 01, 2019
  
 
=External Links=
 
=External Links=
Line 113: Line 153:
 
* [https://github.com/LeelaChessZero/lc0/wiki/FAQ FAQ · LeelaChessZero/lc0 Wiki · GitHub]
 
* [https://github.com/LeelaChessZero/lc0/wiki/FAQ FAQ · LeelaChessZero/lc0 Wiki · GitHub]
 
* [https://github.com/LeelaChessZero/lc0/wiki/Technical-Explanation-of-Leela-Chess-Zero Technical Explanation of Leela Chess Zero · LeelaChessZero/lc0 Wiki · GitHub]
 
* [https://github.com/LeelaChessZero/lc0/wiki/Technical-Explanation-of-Leela-Chess-Zero Technical Explanation of Leela Chess Zero · LeelaChessZero/lc0 Wiki · GitHub]
 +
* [https://github.com/LeelaChessZero/lc0/graphs/contributors Contributors to LeelaChessZero/lc0 · GitHub]
 
* [https://github.com/mooskagh/lc0 GitHub - mooskagh/lc0: The rewritten engine, originally for cudnn. Now all other backends have been ported here]
 
* [https://github.com/mooskagh/lc0 GitHub - mooskagh/lc0: The rewritten engine, originally for cudnn. Now all other backends have been ported here]
 +
* [https://github.com/dkappe/leela-chess-weights/wiki/Distilled-Networks Distilled Networks · dkappe/leela-chess-weights Wiki · GitHub]
 
* [https://blog.lczero.org/ Leela Chess Zero - Blog]
 
* [https://blog.lczero.org/ Leela Chess Zero - Blog]
 
* [https://groups.google.com/forum/#!forum/lczero LCZero – Google Groups]
 
* [https://groups.google.com/forum/#!forum/lczero LCZero – Google Groups]
Line 120: Line 162:
 
* [https://en.chessbase.com/post/leela-chess-zero-alphazero-for-the-pc Leela Chess Zero: AlphaZero for the PC] by [[Albert Silver]], [[ChessBase|ChessBase News]], April 26, 2018
 
* [https://en.chessbase.com/post/leela-chess-zero-alphazero-for-the-pc Leela Chess Zero: AlphaZero for the PC] by [[Albert Silver]], [[ChessBase|ChessBase News]], April 26, 2018
 
* [https://www.youtube.com/watch?v=Crwg2oT9KWE Leela reacts beautifully to Stockfish's outrageous opening greed] by [https://www.youtube.com/channel/UCDUDDmslypVXYoUsZafHSUQ kingscrusher], January 05, 2019, [https://en.wikipedia.org/wiki/YouTube YouTube] Video  
 
* [https://www.youtube.com/watch?v=Crwg2oT9KWE Leela reacts beautifully to Stockfish's outrageous opening greed] by [https://www.youtube.com/channel/UCDUDDmslypVXYoUsZafHSUQ kingscrusher], January 05, 2019, [https://en.wikipedia.org/wiki/YouTube YouTube] Video  
 +
* [http://www.chessdom.com/interview-with-alexander-lyashuk-about-the-recent-success-of-lc0/ Interview with Alexander Lyashuk about the recent success of Lc0], [[Chessdom]], February 6, 2019 » [[TCEC Season 14]]
 +
* [https://en.chessbase.com/post/standing-on-the-shoulders-of-giants Standing on the shoulders of giants] by [[Albert Silver]], [[ChessBase|ChessBase News]], September 18, 2019
 
==Misc==
 
==Misc==
 
* [https://en.wikipedia.org/wiki/Leela Leela from Wikipedia]
 
* [https://en.wikipedia.org/wiki/Leela Leela from Wikipedia]
Line 126: Line 170:
 
* [https://en.wikipedia.org/wiki/Leela_(Doctor_Who) Leela (Doctor Who) from Wikipedia]
 
* [https://en.wikipedia.org/wiki/Leela_(Doctor_Who) Leela (Doctor Who) from Wikipedia]
 
* [https://en.wikipedia.org/wiki/Leela_(Futurama) Leela (Futurama) from Wikipedia]
 
* [https://en.wikipedia.org/wiki/Leela_(Futurama) Leela (Futurama) from Wikipedia]
* [[:Category:Marc Ribot|Marc Ribot's]] Ceramic Dog - Lies My Body Told Me (Live on [https://en.wikipedia.org/wiki/KEXP-FM KEXP], July 20, 2016), [https://en.wikipedia.org/wiki/YouTube YouTube] Video  
+
* [[:Category:Weather Report|Weather Report]] - [https://en.wikipedia.org/wiki/Live_in_Tokyo_(Weather_Report_album) Vertical Invader / T.H. / Dr. Honoris Causa], album [https://en.wikipedia.org/wiki/I_Sing_the_Body_Electric_(album) I Sing the Body Electric] (1972), [https://en.wikipedia.org/wiki/YouTube YouTube] Video  
: {{#evu:https://www.youtube.com/watch?v=YohWv7kIp5o|alignment=left|valignment=top}}
+
: {{#evu:https://www.youtube.com/watch?v=TTJwf4cCUwI|alignment=left|valignment=top}}
  
 
=References=  
 
=References=  
Line 135: Line 179:
 
[[Category:Open Source]]
 
[[Category:Open Source]]
 
[[Category:GPL]]
 
[[Category:GPL]]
[[Category:Marc Ribot]]
+
[[Category:GPU]]
 +
[[Category:NN]]
 +
[[Category:PC]]
 +
[[Category:Windows]]
 +
[[Category:Linux]]
 +
[[Category:Mac]]
 
[[Category:Fiction]]
 
[[Category:Fiction]]
 
[[Category:Given Name]]
 
[[Category:Given Name]]
 +
[[Category:Weather Report]]
 +
[[Category:Wayne Shorter]]
 +
[[Category:Joe Zawinul]]
 +
[[Category:Miroslav Vitouš]]
 +
[[Category:Dom Um Romão]]

Revision as of 17:45, 19 February 2020

Home * Engines * Leela Chess Zero

Lc0 logo [1]

Leela Chess Zero, (LCZero, lc0)
an adaption of Gian-Carlo Pascutto's Leela Zero Go project [2] to Chess, initiated and announced by Stockfish co-author Gary Linscott, who was already responsible for the Stockfish Testing Framework called Fishtest. Leela Chess is open source, released under the terms of GPL version 3 or later, and supports UCI. The goal is to build a strong chess playing entity following the same type of deep learning along with Monte-Carlo tree search (MCTS) techniques of AlphaZero as described in DeepMind's 2017 and 2018 papers [3] [4] [5], but using distributed training for the weights of the deep convolutional neural network (CNN, DNN, DCNN).

Lc0

Leela Chess Zero consists of an executable to play or analyze games, initially dubbed LCZero, soon rewritten by a team around Alexander Lyashuk for better performance and then called Lc0 [6] [7]. This executable, the actual chess engine, performs the MCTS and reads the self-taught CNN, which weights are persistent in a separate file. Lc0 is written in C++14 and may be compiled for various platforms and backends. Since deep CNN approaches are best suited to run massively in parallel on GPUs to perform all the floating point dot products for thousands of neurons, the preferred target platforms are Nvidia GPUs supporting CUDA and CuDNN libraries [8]. Ankan Banerjee wrote the CuDNN backend code, also shared by Deus X and Allie [9]. None CUDA compliant GPUs (AMD) are supported through OpenCL, while much slower pure CPU binaries are possible using BLAS, target systems with or without a graphics card (GPU) are Linux, Mac OS and Windows computers, or BLAS only the Raspberry Pi.

Description

Like AlphaZero, Lc0's evaluates positions using non-linear function approximation based on a deep neural network, rather than the linear function approximation as used in classical chess programs. This neural network takes the board position as input and outputs position evaluation (QValue) and a vector of move probabilities (PValue, policy). Once trained, these network is combined with a Monte-Carlo Tree Search (MCTS) using the policy to narrow down the search to high­probability moves, and using the value in conjunction with a fast rollout policy to evaluate positions in the tree. The MCTS selection is done by a variation of Rosin's UCT improvement dubbed PUCT (Predictor + UCT).

Board Representation

Lc0's color agnostic board is represented by five bitboards (own pieces, opponent pieces, orthogonal sliding pieces, diagonal sliding pieces, and pawns including en passant target information coded as pawns on rank 1 and 8), two king squares, castling rights, and a flag whether the board is color flipped. Getting individual piece bitboards requires some setwise operations such as intersection and set theoretic difference [10].

Network

While AlphaGo used two disjoint networks for policy and value, AlphaZero as well as Leela Chess Zero, share a common "body" connected to disjoint policy and value "heads". The “body” consists of spatial 8x8 planes, using B residual blocks with F filters of kernel size 3x3, stride 1. So far, model sizes FxB of 64x6, 128x10, 192x15, and 256x20 were used.

Concerning nodes per second of the MCTS, smaller models are faster to calculate than larger models. They are faster to train and will earlier recognize progress, but they will also saturate earlier so that at some point more training will no longer improve the engine. Larger and deeper network models will improve the receptivity, the amount of knowledge and pattern to extract from the training samples, with potential for a stronger engine. As a further improvement, Leele Chess Zero applies the Squeeze and Excite (SE) extension to the residual block architecture [11] [12]. The body is connected to both the policy "head" for the move probability distribution, and the value "head" for the evaluation score aka winning probability of the current position and up to seven predecessor positions on the input planes.

Training

Like in AlphaZero, the Zero suffix implies no other initial knowledge than the rules of the game, to build a superhuman player, starting with truly random self-play games to apply reinforcement learning based on the outcome of that games. However, there are derived approaches, such as Albert Silver's Deus X, trying to take a short-cut by initially using supervised learning techniques, such as feeding in high quality games played by other strong chess playing entities, or huge records of positions with a given preferred move. The unsupervised training of the NN is about to minimize the L2-norm of the mean squared error loss of the value output and the policy loss. Further there are experiments to train the value head against not the game outcome, but against the accumulated value for a position after exploring some number of nodes with UCT [13].

The distributed training is realized with an sophisticated client-server model. The client, written entirely in the Go programming language, incorporates Lc0 to produce self-play games.Controlled by the server, the client may download the latest network, will start self-playing, and uploading games to the server, who on the other hand will regularly produce and distribute new neural network weights after a certain amount of games available from contributors. The training software consists of Python code, the pipeline requires NumPy and TensorFlow running on Linux [14]. The server is written in Go along with Python and shell scripts.

Structure Diagrams

Lc0diagram.png

Related to TCEC clone discussions concerning Deus X and Allie aka AllieStein,
Alexander Lyashuk published diagrams with all components of the affected engines,
The above shows the common legend, and the structure of all Leela Chess Zero's components based on current Lc0 engine [15]

Lczero.png

Same diagram, but initial LCZero engine, which played TCEC Season 12 [16]

See also

UCT
PUCT

Forum Posts

2018

Re: Announcing lczero by Daniel Shawul, CCC, January 21, 2018 » Rollout Paradigm
LCZero update (2) by Rein Halbersma, CCC, March 25, 2018
Re: TCEC season 13, 2 NN engines will be participating, Leela and Deus X by Gian-Carlo Pascutto, CCC, August 03, 2018
Re: Has Silver written any code for "his" ZeusX? by Alexander Lyashuk, LCZero Forum, August 02, 2018
Re: How good is the RTX 2080 Ti for Leela? by Ankan Banerjee, CCC, September 16, 2018
Re: How good is the RTX 2080 Ti for Leela? by Ankan Banerjee, CCC, September 17, 2018 [18]
Re: How good is the RTX 2080 Ti for Leela? by Ankan Banerjee, CCC, October 28, 2018
Re: How good is the RTX 2080 Ti for Leela? by Ankan Banerjee, CCC, November 15, 2018
Re: 2900 Elo points progress, 10 million games, 330 nets by crem, CCC, November 25, 2018

2019

Re: Lc0 Evaluation Explanation by Alexander Lyashuk, CCC, September 03, 2019

Blog Posts

2018

Lessons from AlphaZero: Connect Four by Aditya Prasad, Oracle Blog, June 13, 2018
Lessons from AlphaZero (part 3): Parameter Tweaking by Aditya Prasad, Oracle Blog, June 20, 2018
Lessons From AlphaZero (part 4): Improving the Training Target by Vish Abrams, Oracle Blog, June 27, 2018
Lessons From Alpha Zero (part 5): Performance Optimization by Anthony Young, Oracle Blog, July 03, 2018
Lessons From Alpha Zero (part 6) — Hyperparameter Tuning by Anthony Young, Oracle Blog, July 11, 2018

2019

External Links

Chess Engine

Misc

References

  1. Leela Chess Zero (@LeelaChessZero) | Twitter
  2. GitHub - gcp/leela-zero: Go engine with no human-provided knowledge, modeled after the AlphaGo Zero paper
  3. David Silver, Thomas Hubert, Julian Schrittwieser, Ioannis Antonoglou, Matthew Lai, Arthur Guez, Marc Lanctot, Laurent Sifre, Dharshan Kumaran, Thore Graepel, Timothy Lillicrap, Karen Simonyan, Demis Hassabis (2017). Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm. arXiv:1712.01815
  4. David Silver, Thomas Hubert, Julian Schrittwieser, Ioannis Antonoglou, Matthew Lai, Arthur Guez, Marc Lanctot, Laurent Sifre, Dharshan Kumaran, Thore Graepel, Timothy Lillicrap, Karen Simonyan, Demis Hassabis (2018). A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play. Science, Vol. 362, No. 6419
  5. AlphaZero paper, and Lc0 v0.19.1 by crem, LCZero blog, December 07, 2018
  6. lc0 transition · LeelaChessZero/lc0 Wiki · GitHub
  7. Re: TCEC season 13, 2 NN engines will be participating, Leela and Deus X by Gian-Carlo Pascutto, CCC, August 03, 2018
  8. NVIDIA cuDNN | NVIDIA Developer
  9. Re: My failed attempt to change TCEC NN clone rules by Adam Treat, CCC, September 19, 2019
  10. lc0/board.h at master · LeelaChessZero/lc0 · GitHub
  11. Technical Explanation of Leela Chess Zero · LeelaChessZero/lc0 Wiki · GitHub
  12. Squeeze-and-Excitation Networks – Towards Data Science by Paul-Louis Pröve, October 17, 2017
  13. Lessons From AlphaZero (part 4): Improving the Training Target by Vish Abrams, Oracle Blog, June 27, 2018
  14. GitHub - LeelaChessZero/lczero-training: For code etc relating to the network training process.
  15. My failed attempt to change TCEC NN clone rules by Alexander Lyashuk, CCC, September 14, 2019 » TCEC
  16. My failed attempt to change TCEC NN clone rules by Alexander Lyashuk, CCC, September 14, 2019 » TCEC
  17. GeForce 20 series from Wikipedia
  18. Multiply–accumulate operation - Wikipedia
  19. GeForce RTX 2070 Graphics Card | NVIDIA

Up one level