Since you’re here...
... we have a small favour to ask. More people, like you, are reading and supporting our blog: "Chess Engines Diary". And unlike many other sites and blogs, we made the choice to keep our articles open for all, regardless of where they live or what they can afford to pay. We hope you will consider supporting us today. We need your support to continue to exist, because good entries are more and more work time. Every reader contribution, however big or small, is so valuable. Support "Chess Engines Diary" even a small amount– and it only takes a minute. Thank you.
============================== My email: jotes@go2.pl
New version chess engine: Lc0 v.0.31
June 17, 2024
LcZero - CEDR Rating=3698
We have an official version of the LcZero engine. We will start the first tests today!
v0.31.0:
In this version: The blas, cuda, eigen, metal and onnx backends now have support for multihead network architecture and can run BT3/BT4 nets. Updated the internal Elo model to better align with regular Elo for human players. There is a new XLA backend that uses OpenXLA compiler to produce code to execute the neural network. See https://github.com/LeelaChessZero/lc0/wiki/XLA-backend for details. Related are new leela2onnx options to output the HLO format that XLA understands. There is a vastly simplified lc0 interface available by renaming the executable to lc0simple. The backends can now suggest a minibatch size to the search, this is enabled by --minibatch-size=0 (the new default). If the cudnn backend detected an unsupported network architecture it will switch to the cuda backend. Two new selfplay options enable value and policy tournaments. A policy tournament is using a single node policy to select the move to play, while a value tournament searches all possible moves at depth 1 to select the one with the best q. While it is easy to get a single node policy evaluation (go nodes 1 using uci), there was no simple way to get the effect of a value only evalaution, so the --value-only option was added. Button uci options were implemented and a button to clear the tree was added (as hidden option). Support for the uci go mate option was added. The rescorer can now be built from the lc0 code base instead of a separate branch. A dicrete onnx layernorm implementation was added to get around a onnxruntime bug with directml - this has some overhead so it is only enabled for onnx-dml and can be switched off with the alt_layernorm=false backend option. The --onnx2pytoch option was added to leela2onnx to generate pytorch compatible models. There is a cuda min_batch backend option to reduce non-determinism with small batches. New options were added to onnx2leela to fix tf exported onnx models. The onnx backend can now be built for amd's rocm. Fixed a bug where the Contempt effect on eval was too low for nets with natively higher draw rates. Made the WDL Rescale sharpness limit configurable via the --wdl-max-s hidden option. The search task workers can be set automatically, to either 0 for cpu backends or up to 4 depending on the number of cpu cores. This is enabled by --task-workers=-1 (the new default). Changed cuda compilation options to use -arch=native or -arch=all-major if no specific version is requested, with fallback for older cuda that don't support those options. Updated android builds to use openblas 0.3.27. The WDLDrawRateTarget option now accepts the value 0 (new default) to retain raw WDL values if WDLCalibrationElo is set to 0 (default). Improvements to the verbose move stats if `WDLEvalObjectivity is used. The centipawn score is displayed by default for old nets without WDL output. Several assorted fixes and code cleanups.
Individual statistics:Lc0 BT4
Clover 6.1 5.5/8 +3 8 Games RofChade 3.1 4.5/8 +1 8 Games Dragon 3.3 4.5/8 +1 8 Games SF-PRO 19.11.2023 4/8 +0 8 Games YuliGM PrO 15 3.5/8 -1 8 Games ShashChess 34.5 3.5/8 -1 8 Games Sawfish 2TC 3.5/6 +1 6 Games Polyfish 231120 3.5/6 +1 6 Games Hazard 4.1 3/6 +0 6 Games Raid 3.1 3/6 +0 6 Games Eman 9.80 3/6 +0 6 Games Incognito 4 3/6 +0 6 Games Vanilla 14c 3/6 +0 6 Games Patzer AI X-256 2.5/6 -1 6 Games SugaR AI SE 2.5/5 +0 5 Games Altair 6.0.0 3.5/4 +3 4 Games Caissa 1.15 3/4 +2 4 Games Seer 2.8.0 2.5/4 +1 4 Games Fisherov chess monk 1.2 2.5/4 +1 4 Games Crystal 7 CMB 2.5/4 +1 4 Games Chess-System-Tal-2 2.5/4 +1 4 Games
github:https://github.com/LeelaChessZero/lc0/releases/
Labels
LCZero
Comments
Post a Comment
Popular posts from this blog
New Strong Engines Test, by Chess Engines Diary, 2024.04.12
April 13, 2024
💾 552 (!) games from the tournament download 👍@chessenginesdiary Country - Poland, City - Malbork 🕓 Time 3'+3" 💻HP Pavilion i5-1035G1 8GB RAM🖬 GUI-Banksia All CEDR 317.321 games download (01.04.2024 - 3'+3") HypnoS 030424, Yuli GM Pro 16, Stockfish 16.1 and ShashChess 35 . These four engines scored the same number of points and placed at the top of the table in this strong tournament. Tech table: Engine KN/move NPS dep/mov time/mov mov/game time/game fails Alexandria 6.1.0 3218 576429 30.2 5.6 56.7 316.2 Berserk 13 4779 841025 36.7 5.7 56.0 318.1 Brainlearn 28 2275 366977 31.4 6.2 43.5 269.4 Caissa 1.18 5492 923486 30.2 5.9 50.4 299.7 Clover 6.1.19 4803 854377 31.4 5.6 62.2 349.4 Cool Iris 12.10 2201 356730 27.6 6.2 51.2 315.5 CorChess 20240331 2253 372774 27.4 6.0 52.2 315.8 Crystal 8 3582 571994 20.3 6.3 47.7 298.8 Fire 9.3 5225 876565 22.0 6.0 55.5 331.1 Fisherov chess monk 1.2 3487 619728 36.0 5.6 60.4 340.0 Hyp
Post a Comment
Read more
End of the Torch Experiement?
June 19, 2024
Andrew Grant: "Almost exactly one year ago we announced the release of Torch. We were a bit ambitious at the time, and announced it as the #2 engine, after beating Leela in the most recent Bullet event at CCC at the time. Today, that claim rings much more true, although still has some minor points of contention. Torch has beaten Leela in multiple Bullet and Blitz events in a row; Never managed to win one of the Rapid events; But has beaten Leela in a h2h classical event. Chesscom got into the space out of a need to develop a replacement for Komodo. They needed a strong chess engine which could be used for integrating their other products. Many features on Chesscom, from Game Review, to puzzle generation, to Bots (obviously!) are powered by high quality engines. The goal was always for that. In fact, despite me being the only full time employee working on Torch, a plurality of my time was spent on things not related to Torch's strength. Its a bit self-aggrandizing, but the Tor
Post a Comment
Read more
SugaR AI SE wins New Strong Engines Test, by Chess Engines Diary, 2024.05.25-29
May 30, 2024
The tournament of strong chess engines was won by SugaR AI SE . This is a very good compilation though. In places 2 to 4: Stockfish 20240523 Ivec , Sun 15 and Deep Blue-20240518-assassin . All CEDR 333.448 games download (01.05.2024 - 3'+3") 💾 600 (!) games from the tournament download 👍 @chessenginesdiary Country - Poland, City - Malbork 🕓 Time 3'+3" 💻HP Pavilion i5-1035G1 8GB RAM🖬 GUI-Banksia Tech table: Engine KN/mov NPS dep/mov time/mov mov/game time/game fail Caissa 1.18 4920 906675 27.5 5.4 64.4 349.7 Chess-System-Tal-2.00 2263 448276 20.6 5.0 64.5 325.8 Clover 6.1.24 JA 4435 820982 27.6 5.4 70.4 380.3 Cool Iris 12.10 2138 356907 26.7 6.0 54.2 324.7 CorChess 20240523 1885 330718 28.5 5.7 55.7 317.3 Crystal 8 3408 585347 20.7 5.8 55.3 321.7 Deep Blue-20240518 1717 308790 31.1 5.6 59.6 331.4 Dragon 1 by Komodo 3688 693178 24.0 5.3 68.2 362.7 Fisherov chess monk 1.2 3290 620116 34.0 5.3 67.9 360.2 HypnoS 160524 1871 328
Post a Comment
Read more
New chess engine: Steel Fighter (Stockfish derivatives)
May 27, 2024
Steel Fighter by Solista. Eduard NemethI created this engine for playing on Lichess. The engine uses an extremely fast random op. MultiPV mode! With a value of 2 (default) the engine is only slightly slower than Stockfish dev. Positionally, the Enghine plays a little weaker than Stockfish dev, but tactically it is much better. On Lichess, where half of the enemies are non-Stockfish and make more tactical mistakes, you'll have more fun with Steel Fighter. Steel Fighter is also a good addition for short tactical analyses. Download Win 64 Bit (avx2, bmi2, sse41, source code): https://pixeldrain.com/u/Yqp3g6TT Note: This engine uses special code and specially adapted networks. A change leads to worse results and also incorrect ratings. Steel Fighter download
Post a Comment
Read more
Chess engine: Deep Ripper -1 (Stockfish derivatives)
June 13, 2024
Deep Ripper (Stockfish derivatives) byEduard Nemeth Author: " Here is the first official version of Deep Ripper. For now, I have decided on a version that can be used for any hardware. If you only have a few cores, you should set the number of variants to 1 or 2. However, 4 variants are standard (1 to 10 can be set), which deliver very good analysis results on 6 or more cores. Playing on servers is also possible. I hope you enjoy testing." Deep Ripper - 1 download Deep Ripper - 1 download This engine uses the Random Op. MultiPV mode. Here, several variants are examined in parallel and, after e.g. 16 plies, merged into a single variant. This is not the same as the normal multi-variant mode, where all variants are examined separately. The results can be completely different! So it makes no sense to create an identical engine without the random mode.
Post a Comment
Read more
