sabakihq / leelasabaki Goto Github PK

Hi, I have everything working except I don't know how to get the %WR below the game tree.
My argument is currently:
--heatmap D:\Downloads\leela-zero-0.15-win64\leela-zero-0.15-win64\leelaz.exe -w D:\Downloads\leela-zero-0.15-win64\leela-zero-0.15-win64\networks\b69158ea816e75b7ffea06142d90d2b1199b48265e75e292f3d5bd3c081d78ee --playouts 100 --noponder
Also curious to see full list of arguments I can use. Thanks in advance.

Can we open a sgf file, and then let LeelaSabaki generate variations for specified moves?

As a freshman user of leela, I would like to see the heatmap every step I play. however I must type "heatmap" at the bottom of the software then it appears the heatmap. Another thing is that whether I can see the probability of win in the chessboard especially those optional sites.

In the leelasabaki directions: it says "Arguments: --heatmap my_path/leelaz -w my_path/weights.txt --playouts 100 --noponder" . But i cannot find the weights.txt file.

Tested several different ways and does not launch.

I used command ./leelasabaki /home/palm/leela-zero/build/leelaz and got /home/palm/leela-zero/build/leelaz: /usr/local/cuda/lib64/libOpenCL.so.1: no version information available (required by /home/palm/leela-zero/build/leelaz).
But if I use ./leelasabaki /home/palm/leela-zero/build/leelaz --weights /home/palm/leela-zero/build/best-network --gtp, it just like a normal leela when I run in command

Normal Sabaki works fine in my machine.

Sorry for my gramma

Just like Lizzie

LeelaSabaki doesn't seem fast enough to relay responses from Leela sometimes.

Hi,

I was running a match between two different weightfiles using the same LeelaSabaki and LeelaZ executables (just different command line options in the engine management). After about 100 moves, Sabaki freezes and crashes with this popup:

One thing I noticed as well is that both players keep decreasing their "Thinking at most XX seconds" at every move. It starts at 50.0 and then counts down by 0.2s with the small weightfiles (5x64) and by 0.4s per move with the large weightfile (6x128 supervised). Not sure if that has anything to do with it, I tried with lower playouts so they don't get clipped by time; the crash still happens.

I cant get leela zero to work with sabaki.
This is my config:

This is the error I am getting:

Leela seems to be working fine running from console.

how many options are available? what's the meaning of --playouts option?

When invoking leelasabaki, either from the CLI or through Sabaki, it immediately crashes with:

/snapshot/LeelaSabaki/src/index.js:95
lineReader.on('line', async input => {
                      ^^^^^

SyntaxError: missing ) after argument list
    at createScript (vm.js:56:10)
    at Object.runInThisContext (vm.js:97:10)
    at Module._compile (module.js:542:28)
    at Module._compile (pkg/prelude/bootstrap.js:1250:32)
    at Object.Module._extensions..js (module.js:579:10)
    at Module.load (module.js:487:32)
    at tryModuleLoad (module.js:446:12)
    at Function.Module._load (module.js:438:3)
    at Module.runMain (pkg/prelude/bootstrap.js:1298:12)
    at run (bootstrap_node.js:419:7)

I don't know anything about the NodeJS ecosystem, but this smells like some sort of language version mismatch to me. I'm running:

$ node --version
v6.10.3
$ npm --version
3.10.10

It could be interesting to store the heapmap in the SGF. That will allow the see the heatmap in the replay of a game for analysis.

Most chess engines have the "analyze" command, meaning the engine gives you all the variations it visits and the corresponding win%s, but it doesn't make a move on the board. That's very conveninent when analyzing games because you don't add nodes that you don't want to the game tree.

It'd be nice to have an option for leelasabaki to behave like that, so the F5 key calls genmove, but leelasabaki doesn't put the final stone on the board.

rt

internal/child_process.js:323
throw errnoException(err, 'spawn');
^
Error: spawn EACCES
at _errnoException (util.js:1024:11)
at ChildProcess.spawn (internal/child_process.js:323:11)
at Object.exports.spawn (child_process.js:502:9)
at childProcess.spawn (pkg/prelude/bootstrap.js:1392:27)
at new GTPEngine (/snapshot/LeelaSabaki/src/GTPEngine.js:10:24)
at Object. (/snapshot/LeelaSabaki/src/index.js:36:14)
at Module._compile (pkg/prelude/bootstrap.js:1243:22)
at Object.Module._extensions..js (module.js:644:10)
at Module.load (module.js:552:32)
at tryModuleLoad (module.js:495:12)

I am having this issue at startup of LeelaSabaki using

./leelasabaki --heatmap path/leela-zero/src -w path/leela-zero/weights.txt --playouts 100 --noponder

My node and npm are installed using Homebrew.

Error message is shown when trying to play with leelaSabaki:

"Current board arrangement can’t be recreated on the GTP engine."

Engine is called with:
path: my_path/leelasabaki-linux
arguments: my_path/leelaz --gtp -w my_path/weights.txt --playouts 10 --noponder

Playing with leelaz without leelasabaki-linux works fine. Using sabaki-v0.33.2-linux-x64.AppImage with leelsabaki-v0.3.0-linux-macos-win-x64.7z under ubuntu.

What am I doing wrong?

The 1st and 2nd recommended moves seem too big and splotchy for my eyes. I vote to make them the same size as the 3rd move (C12 in image) and then use colors to distinguish them (Leela uses: red, yellow, and blue.)

While the engine manager in Sabaki is nice and allow flexibility in the use of different engines, it would be nice to have a default one directly integrated into Sabaki.

Maybe we could find a way to add new engines trough npm dependency. That would be an alternative to the current binary command line based engines.

The npm plugin would expose a JS class with known methods to be called during the use of the engine.

I was wondering if I'm doing something wrong -- I'm running the latest Sabaki to which I've added the latest CLI Leela engine (v0.11) which seems to work fine, it thinks a bit and generates moves, plus the game tree variations.

However I have no idea how to make heatmap work, is there a special theme or any special thing I need to do in Sabaki? (other than passing --heatmap to leelasabaki which I already do).

sabaki not working, stupid command line syntax issues, sabaki leela is convoluted pile of $*!%, sabaki has to interface with sabaki-leela, then sabaki-leela has to interfac with node.js, then that interface with leelz which interface with the network weights... , unnecessarily complicated steps, Sabaki shoudl just work with leelaz out the box, we have two different versions of leela, four different versions of alphago, and two different versions of sabaki, three different network sizes, even the example syntax doesn't make it clear if it could be / or \ and if the thing could be on single or double quotes I mean wtf

Hi,

I've made some changes in my fork of leelaSabaki.
https://github.com/5525345551/LeelaSabaki

• It passes the ponder variations to sabaki for 'play' as well as 'genmove'
• Can choose to include 'black' or 'white' variations only
• limit the variation depth

These are useful to use when reviewing a game.

Specify your player color e.g. '--white' and it will only include that colour, or '--white --black' to get both. For the non bot variations it uses the ponder console output.

I like to use '--flat' and '--limitdepth' so variations are 7 moves deep and can see it all easily.

I tried to have a '--depth n' moves but it will be difficult because of leelapath index is first non '-' arg.

This also needs a patch (dirty hack) to Sabaki, see my github fork of sabaki. https://github.com/5525345551/Sabaki
3 lines of code added.
Although it is buggy when stopping and starting a game because the sabaki move state is messed up. Sorry i dont know how to fix it better. :(

It would be good to put labels node.[LB] for variation first moves 'A', 'B', 'C', etc in the main branch of the game tree when the variations are added, but it required more hacks to sabaki.

Perhaps add this sgf 'LB' info to the JSON 'genmovelog' and then add that to the tree position node and board.markup in Sabaki?

Please take a look to see if you would like a pull request, but it is a bad hack :)

Hi, yichuan
I want to run the LeelaSabaki on my Mac-pro, but the process LeelaSabaki and LeelaZero crashed about 5 seconds later.
Can you help me? Thank you very much !

1、LeelaSabaki v0.3.0
2、Leela Zero 0.15
3、macOS High Sierra (10.13.6 )
4、MacBook Pro (13-inch, 2018, Four Thunderbolt 3 Ports)
5、/data0/leela-zero/leelasabaki-macos --flat --heatmap /data0/leela-zero/src/leelaz -t 4 -w /data0/leela-zero/best-network --playouts 10 --noponder

Using 4 thread(s).
RNG seed: 9153934374649917980
Detecting residual layers...v1...256 channels...40 blocks.
Initializing OpenCL.
Detected 1 OpenCL platforms.
Platform version: OpenCL 1.2 (May 24 2018 22:33:53)
Platform profile: FULL_PROFILE
Platform name: Apple
Platform vendor: Apple
Device ID: 0
Device name: Intel(R) Core(TM) i5-8259U CPU @ 2.30GHz
Device type: CPU
Device vendor: Intel
Device driver: 1.1
Device speed: 2300 MHz
Device cores: 8 CU
Device score: 512
Device ID: 1
Device name: Intel(R) Iris(TM) Plus Graphics 655
Device type: GPU
Device vendor: Intel Inc.
Device driver: 1.2(Aug 14 2018 19:16:17)
Device speed: 1050 MHz
Device cores: 48 CU
Device score: 612
Selected platform: Apple
Selected device: Intel(R) Iris(TM) Plus Graphics 655
with OpenCL 1.2 capability.
Loaded existing SGEMM tuning.
Wavefront/Warp size: 8
Max workgroup size: 256
Max workgroup dimensions: 256 256 256

bash-3.2# ps aux|grep -i leela
root 3572 0.0 0.0 4267768 908 s001 S+ 9:36下午 0:00.00 grep -i leela
bash-3.2#