Ceres Chess Engine

[Nemesis]

I guess that was a mere fluctuation; I've checked that postion and it's 0.10~0.00 at depth 41.

Thanks Mr Goi - appreciate the check !!

[necunoscutul1975]

good moprning friend io found ceres chess engine but not work on my windows gtx1050 ti

can some hel me ? here is te old ceres file ?


https://u.pcloud.link/publink/show?code=XZO4KQXZekjolEiQzpjcVOHWr8X1jbuaKmU7

MOD >> do NOT duplicate OK.

Read this thread from the beginning - there's a set of instructions that you need to follow that I quoted (from Ribbit) - have you done so ?

[timetraveller]

Guidance Note
by Stefan Pohl

Boxed version of Ceres 0.8 is here (huge download!):
https://drive.google.com/file/d/1yPF13m7rgsP9xM6f1UsOSeLhS2_lW1gz/view?usp=sharing

Unzip the download and you get a folder: Ceres. In there are 2 folders:
lc0_dev, runtimes. I put the net, Ceres should use in the lc0_dev folder.
In the Ceres-folder, there is the configuration-file, called Ceres.json (see below) and
the Ceres-binary (Ceres.exe)

Before you can use Ceres, you have to install DotNet5 runtime environment by Microsoft:
https://dotnet.microsoft.com/download/dotnet/thank-you/runtime-desktop-5.0.1-windows-x64-installer

And you have to install Cuda 11.2:
https://developer.nvidia.com/cuda-downloads

In the Ceres.json file (Ceres folder), the net has to specified
(otherwise default net 703810 is used). My file looks like this:

{
"DirLC0Binaries": ".",
"DirLC0Networks": "C:\\Cutechess\\Engines\\Ceres\\lc0_dev",
"DirEPD": ".",
"DirPGN": ".",
"DirCeresOutput": ".",
"DirExternalEngines": ".",
"DirTablebases": "",
"DefaultNetworkSpecString": "LC0:66680",
"DefaultDeviceSpecString": "GPU:0",
"URLLC0Networks": "http://training.lczero.org/networks";,
"DebugAllowed": false,
"LaunchMonitor": false,
"LogInfo": false,
"LogWarn": false
}

Line "DirLC0Networks" must contain the full path to the net. And DefaultNetworkSpecString is the
name. BUT the netfile name on my HDD is (in this example) 66680.pb.gz (!!!) But the extension
(.pb.gz) must not be written here. And "LC0:" must be added. Strange. But works. Make sure, your
netfile-name ends with .pb.gz And (I read on discord) make sure, the netfilename contains numbers,
only, before the .pb.gz extension (!!!) (66680.pb.gz is fine, for example).

Of course your NVIDIA RTX card driver has to be updated to the latest version, use the
GeForce Experience program or visit the NVIDIA website and download the driver manually.

Reboot the system after installing and updating all that stuff, before starting Ceres!!!
Then try to start Ceres.exe with doubleclick in Windows and type "go infinite", then Ceres
should start calculating after some seconds. If this happens, you should be able to use Ceres
as Uci-engine in a GUI. I only tried cutechess-cli. Works in there.

[kramnik]

RTX...Does it work with 7xx-8xx-9xx series?

Guidance Note
by Stefan Pohl

Boxed version of Ceres 0.8 is here (huge download!):
https://drive.google.com/file/d/1yPF13m7rgsP9xM6f1UsOSeLhS2_lW1gz/view?usp=sharing

Unzip the download and you get a folder: Ceres. In there are 2 folders:
lc0_dev, runtimes. I put the net, Ceres should use in the lc0_dev folder.
In the Ceres-folder, there is the configuration-file, called Ceres.json (see below) and
the Ceres-binary (Ceres.exe)

Before you can use Ceres, you have to install DotNet5 runtime environment by Microsoft:
https://dotnet.microsoft.com/download/dotnet/thank-you/runtime-desktop-5.0.1-windows-x64-installer

And you have to install Cuda 11.2:
https://developer.nvidia.com/cuda-downloads

In the Ceres.json file (Ceres folder), the net has to specified
(otherwise default net 703810 is used). My file looks like this:

{
"DirLC0Binaries": ".",
"DirLC0Networks": "C:\\Cutechess\\Engines\\Ceres\\lc0_dev",
"DirEPD": ".",
"DirPGN": ".",
"DirCeresOutput": ".",
"DirExternalEngines": ".",
"DirTablebases": "",
"DefaultNetworkSpecString": "LC0:66680",
"DefaultDeviceSpecString": "GPU:0",
"URLLC0Networks": "http://training.lczero.org/networks";,
"DebugAllowed": false,
"LaunchMonitor": false,
"LogInfo": false,
"LogWarn": false
}

Line "DirLC0Networks" must contain the full path to the net. And DefaultNetworkSpecString is the
name. BUT the netfile name on my HDD is (in this example) 66680.pb.gz (!!!) But the extension
(.pb.gz) must not be written here. And "LC0:" must be added. Strange. But works. Make sure, your
netfile-name ends with .pb.gz And (I read on discord) make sure, the netfilename contains numbers,
only, before the .pb.gz extension (!!!) (66680.pb.gz is fine, for example).

Of course your NVIDIA RTX card driver has to be updated to the latest version, use the
GeForce Experience program or visit the NVIDIA website and download the driver manually.

Reboot the system after installing and updating all that stuff, before starting Ceres!!!
Then try to start Ceres.exe with doubleclick in Windows and type "go infinite", then Ceres
should start calculating after some seconds. If this happens, you should be able to use Ceres
as Uci-engine in a GUI. I only tried cutechess-cli. Works in there.

[Nemesis]

Ceres 0.86

ChadToday at 3:16 PM
I've uploaded the latest Ceres code as a boxed build. I haven't tested it so caveat emptor:

https://www.chadhosting.xyz/programs/Ceres_boxed_ebb28ca.exe

[Kater Carlo]

Ceres Release v0.87 is available with interesting improvements, still experimental:
https://github.com/dje-dev/Ceres/releases/tag/0.87

[Nemesis]

Release notes for Ceres 0.87


Previous versions of Ceres had serious difficulties winning won endgames when tablebases were enabled. This problem is now partially corrected (due to DTZ file support) but still needs more work. Therefore it is suggested to use reasonable adjudication rules when running tournaments.

The LC0.DLL used to access the Leela Chess Zero backend code has been updated. Although the prior version will continue to work with 0.87, it will not provide the best experience because it lacks tablebase DTZ support (for finding winning moves in tablebase endings).

The excellent Nibbler GUI interface is now supported due to newly supported UCI setoptions (MultiPV, LogLiveStats and UCI_ShowWDL).

A bug causing incorrect principal variation moves to be output to UCI clients is corrected.

The format of network specification strings (such as "LC0:703810") are now more flexible:

the "LC0:" can optionally be omitted
the network specification (such as "703810") can optionally instead be a complete file path (such as "c:\nn\weights2_703810.pb.gz")
The Ceres.json key "SyzygyPath" now can be used to set the tablebase directories (DirTablebases is retained for backward compatibility).
Optionally multiple semicolon-separated tablebase paths can be specified.

A subset of the setoption commands supported by Leela are now supported:

SmartPruningFactor (only supported for value 0)
SyzygyPath
CPUCT
CPUCTAtRoot
CPUCTBase
CPUCTBaseAtRoot
CPUCTFactor
CPUCTFactorAtRoot
PolicyTemperature
VerboseMoveStats
LogFile
MultiPV
These options (except MultiPV) are also supported in Ceres.json under the same name.

The LogFile option creates a log file containing detailed information about each search, including
the verbose move statistics, a dump of the principal variation and associated statistics for each move,
and information about the reason for the search termination.

In the special case of very small searches (e.g. < 1000 nodes), search speed was significantly improved.

Certain search and time management parameters were tuned further tuned.

Numerous small bugs and incompatibilities were corrected.

The file storage required for build, packaging and deployment is reduced dramatically (from about 500mb to 35mb)
due to elimination of unused Nuget packages.

Play quality is believed improved, with magnitude of improvement varying widely depending on the network and time control,
often falling in the 10 to 20+ Elo range.

Continuing improvements to code clarity, simplicity and documentation were made.

A binary installation package was created. Hopefully it should be possible to run Ceres directly from the unpacked files. Two subdirectories (CUDA_v11.1 and CUDA_v11.2) contain the required LC0.DLL file. Depending on the version of CUDA installed (11.1 or 11.2), one or the other of these files should be placed in the directory containing the Leela Chess Zero version 0.26.3 binaries (and specified in the Ceres.json file as "DirLC0Binaries"). Please note this binary distribution is experimental and not yet widely tested.

https://github.com/dje-dev/Ceres/releases/download/0.87/Ceres_v0.87.zip

https://github.com/dje-dev/Ceres/archive/0.87.zip

[kamikazepawn]

hi folks...it would be really cool if we could have (Ceres) with AMD GPU ... for those who don't have NViDia ... waiting anxiously ... best

[Kater Carlo]

New release of Ceres 0.88 today:

Code: Select all

https://github.com/dje-dev/Ceres/releases/tag/v0.88

Upgrading Ceres typically involves the following steps:
download the ZIP file and unpack files in a new directory
copy forward the Ceres.json successfully used from the prior installation into this directory
overwrite the LC0.DLL in the directory with LC0 binaries (identified in DirLC0Binaries entry in Ceres.json) by copying one of the LC0.DLL files included in this ZIP file (based on the version of CUDA installed on your computer).
Note that the CUDA version of LC0 must have been downloaded into that directory (probably both 0.26.3 and 0.27 will work).

Notable Changes
Compatibility with many clients and GUIs is improved. Extensive testing has been performed against Nibbler and Arena and Cutechess, and also the tournament managers in Cutechess and Arena and cutechess-cli.

[marco277]

Good morning friends and excuse my ignorance but this engine can only be run on Windows 10

[Kater Carlo]

Hardware/Software Requirements:
Windows 10 (a Linux version is under development)
Processor based on Intel/AMD x86 architecture
Processor support for AVX instruction set
At least one NVIDIA GPU with CUDA drivers installed
Minimum 8GB of memory (recommended)

High requirements, indeed!

[marco277]

Gracias Amigo

[timetraveller]

v0.89

Notable Changes:
-Ceres is more aggressive about sometimes choosing the "best Q" move at root (over the classic "best N"), with the incidence rising from approximately 4% to 8%. There is some evidence this improves play quality, but more testing is needed to quantify impact.
-Filenames with embedded spaces are now accepted (e.g. for weights files)
-Network files are autodetected as being in either zipped or unzipped format
-Culture insensitive numeric formats are now adapted as standard (e.g. decimal point with numbers)
-Time management is adjusted to be slightly more adaptive and conservative
-Errors with processing of rapid consecutive UCI go commands have been (at least partly) corrected
-Two UCI setoptions are available for logging: LogFile is for UCI dialog with engine, and SearchLogFile is for detailed internal information on each search conducted (such as verbose move stats and principal variation dump)

Download

[supersharp77]

Ceres CPU....Much Needed..(testing)..Pass It ON!! Thx kc.....

[EzioMagnifico]

v0.90-rc1
Thanks to the Author dje-dev

Download : https://github.com/dje-dev/Ceres/releases

Ceres Release Notes - version 0.90-rc1 (Release Candidate 1).

Community help and feedback in testing v0.90 would be welcomed. Primary areas of enhancement include:

Compatability (Linux support, also GPUs with limited memory)
Installation (LC0.DLL no longer required)
Search speed faster by 10% to 25% (due to enhancements to CUDA backend and MCTS engine), especially on more recent NVIDIA hardware and CUDA 11.3+.
Because of the major backend changes (about 10,000 lines of new C# code) it is likely that issues will be identified relating to untested hardware and software configurations. Please feel free to open an issue if you are encountering difficulties, or post in #help of the Discord channel for Leela Chess Zero.

Internal testing suggests that play strength (with T60) is significantly improved, even on relatively modest hardware (Windows laptop with 2070 GPU) and moderate time controls (such as 60 seconds per game). However independent community assessment is necessary and would be welcomed.

Known minor issues:

loading of networks (at initialization) is a little slow
in multigpu configurations the new feature to reduce GPU memory consumption is disabled
Details

The same binaries support both Windows and Linux operating systems. To run on Linux, use: "dotnet ./Ceres.dll"

A native C# backend for CUDA instead of relying upon an external library LC0.DLL, yielding simplified installation and improved performance. This backend is largely a transliteration of the Leela Chess Zero CUDA backend in C++ (largely by Ankan) into C#
This work leveraged open source ManagedCUDA project (by Michael Kunz) to provide the object-oriented bindings to the CUDA C API.
The implementation also features several enhancements:

use of CUDA graphs to precompile the entire neural network into a single CUDA operation, yielding speedups of 5% to 20%
(greater with smaller networks and smaller batch sizes), especially on more recent hardware and CUDA 11.3 or above.
a supplemental CUDA kernel which reduces GPU/host bandwidth requirements (copies policy only for legal moves)
reduced GPU memory consumption and network load times (by about 25%)
diagnostic and introspection features for developers, such as optionally capturing inner layer timings and activations
integrated native Syzygy tablebase probing (thru 7 man) based upon transliteration of the Fathom library (by Ronald de Man, basil, and Jon Dart) from C++ to C#

search speed optimizations, particularly for long searches on higher-end GPUs. For example, 425k nodes per second is achievable on 2017 vintage CPUs.

some adjustments and tuning to MCTS parameters yielding slightly improved play quality in most situations

two additional command line options (BACKENDBENCH and BENCHMARK) emulating LC0 functionality.
See below for an example of the output from the benchmark command which also compare against LC0 v0.28rc1, e.g.
"Ceres benchmark opponent=lc0 limit=10sm"

various bug fixes, including those involving use of MultiPV feature

ongoing work to improve source code clarity and documentation

ongoing work to make the Ceres code base (set of object-oriented classes for general chess operations, MCTS search, and neural network training) more flexible, performant, and well documented to facilitate research in Chess programming.

------ Windows Laptop with 2070, T75 network, 10 seconds each -------

Ceres Benchmark Results =======
Total time(ms) : 339,721
Nodes searched : 26,618,539
Avg nodes/sec : 78,354
Median nodes/sec : 46,276
Positions faster : 32

LC0 Benchmark Results =========
Total time(ms) : 340,487
Nodes searched : 22,238,029
Avg nodes/sec : 65,312
Median nodes/sec : 41,076
Positions faster : 2