Seer

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

User avatar
Guenther
Posts: 4611
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: Seer

Post by Guenther »

Graham Banks wrote: Sat Nov 14, 2020 11:05 am
Guenther wrote: Sat Nov 14, 2020 11:01 am
Graham Banks wrote: Sat Nov 14, 2020 9:44 am
Guenther wrote: Sat Nov 14, 2020 9:33 am
Graham Banks wrote: Sat Nov 14, 2020 9:28 am

Must be a communication problem between the engine and ChessGUI.
Perhaps somebody else who uses ChessGUI can confirm.
Well, what is your setting in ChessGUI for the Weights file and how does it look?
Image
A last attempt, what happens, if you just add the file name, as it seems the nn file is already located in the same directory?
Normally then it should work w/o the absolute path.
So instead 'C:\ChessGUI_net\ENGINES\Seer 1.1 64-bit\save.bin' simply 'save.bin'
I've tried that as well, Guenther.
Will be interesting to see whether Andreas (or any other ChessGUI user) can get Seer working correctly.
After a very very long time I succeeded running ChessGUI at all.
Here there happen strange things with Seer in ChessGUI (portable 2.45m)

It will always crash here after trying to install it and looking at the options tab
it adds the string(type) as value with one char omitted, instead of nothing.
(The omitted char could be also due to my non standard default font-size,
still string is just the type not the value and should not be there)

Image

Did you install Seer 1.1 through the installation dialogue or by adding it to the engines list?

I have the feeling ChessGUI has a problem with the missing default value in the Weights option.

Code: Select all

uci
id name Seer 1.1
id author C. McMonigle
option name Clear Hash type button
option name Threads type spin default 1 min 1 max 512
option name Hash type spin default 128 min 1 max 65536
option name Weights type string
uciok
https://rwbc-chess.de

trollwatch:
Talkchess nowadays is a joke - it is full of trolls/idiots/people stuck in the pleistocene > 80% of the posts fall into this category...
User avatar
Graham Banks
Posts: 41473
Joined: Sun Feb 26, 2006 10:52 am
Location: Auckland, NZ

Re: Seer

Post by Graham Banks »

Guenther wrote: Sat Nov 14, 2020 12:29 pm
Graham Banks wrote: Sat Nov 14, 2020 11:05 am
Guenther wrote: Sat Nov 14, 2020 11:01 am
Graham Banks wrote: Sat Nov 14, 2020 9:44 am
Guenther wrote: Sat Nov 14, 2020 9:33 am
Graham Banks wrote: Sat Nov 14, 2020 9:28 am

Must be a communication problem between the engine and ChessGUI.
Perhaps somebody else who uses ChessGUI can confirm.
Well, what is your setting in ChessGUI for the Weights file and how does it look?
Image
A last attempt, what happens, if you just add the file name, as it seems the nn file is already located in the same directory?
Normally then it should work w/o the absolute path.
So instead 'C:\ChessGUI_net\ENGINES\Seer 1.1 64-bit\save.bin' simply 'save.bin'
I've tried that as well, Guenther.
Will be interesting to see whether Andreas (or any other ChessGUI user) can get Seer working correctly.
After a very very long time I succeeded running ChessGUI at all.
Here there happen strange things with Seer in ChessGUI (portable 2.45m)

It will always crash here after trying to install it and looking at the options tab
it adds the string(type) as value with one char omitted, instead of nothing.

I noticed that. Perhaps that is the issue. Wonder why it happens.

Image

Did you install Seer 1.1 through the installation dialogue or by adding it to the engines list?

I go to engines/install engine, and install from there.

I have the feeling ChessGUI has a problem with the missing default value in the Weights option.

Code: Select all

uci
id name Seer 1.1
id author C. McMonigle
option name Clear Hash type button
option name Threads type spin default 1 min 1 max 512
option name Hash type spin default 128 min 1 max 65536
option name Weights type string
uciok
gbanksnz at gmail.com
User avatar
Guenther
Posts: 4611
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: Seer

Post by Guenther »

Guenther wrote: Sat Nov 14, 2020 12:29 pm ...

Image

Did you install Seer 1.1 through the installation dialogue or by adding it to the engines list?

I have the feeling ChessGUI has a problem with the missing default value in the Weights option.

Code: Select all

uci
id name Seer 1.1
id author C. McMonigle
option name Clear Hash type button
option name Threads type spin default 1 min 1 max 512
option name Hash type spin default 128 min 1 max 65536
option name Weights type string
uciok
There seem to be several things going on.
At least here Seer will always crash if I try to install it through the dialogue in ChessGUI.

Edit: It will be still saved to the engine list if I save the settings despite the crash.

When installing ChessGUI sends a stop after it has gathered all options
and Seer reacts with bestmove h1h1 and then it crashes after ChessGUI sends quit.

I could replicate this behaviour from Seer in cmd.

Code: Select all

0.000: - Starting server: seer111-64.exe
0.031: - Priority class set to normal
0.047: - Thread priority set to normal
0.047: > uci
1.233: < id name Seer 1.11
1.233: < id author Connor McMonigle
1.233: < option name Clear Hash type button
1.233: < option name Threads type spin default 1 min 1 max 512
1.233: < option name Hash type spin default 128 min 1 max 65536
1.233: < option name Weights type string default save.bin
1.233: < uciok
1.373: > stop
1.373: < bestmove h1h1
1.373: > quit
1.373: < 
1.373: < This application has requested the Runtime to terminate it in an unusual way.
1.373: < Please contact the application's support team for more information.
3.370: - Client closed
3.370: - Stopping server communication
8.377: - Closing log
https://rwbc-chess.de

trollwatch:
Talkchess nowadays is a joke - it is full of trolls/idiots/people stuck in the pleistocene > 80% of the posts fall into this category...
User avatar
Guenther
Posts: 4611
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: Seer

Post by Guenther »

Your problem seems to be fixed though with current dev version:
I played a test game Seer vs. Seer and the option (which has a now a default value) is sent
and it plays with the Weights file!

I could upload my dev version here for verifying for yourself.
It has save.bin set to default by me now. So you need to change nothing if the nn file is in the same directory.

Done:
https://rwbc-chess.de/Downloads/Engines/seer_111-64.7z

Code: Select all

 ================ SendFirstCommands(Seer 1.11), nProtocol = 1, EngineCodeHex = 00000020 ====
SendFirstCommands UCI
SendToEng2Time 0000000168987100 : Eng02 (Seer 1.11) -> uci
.
Scheduling StartEngineTimerEvent() for Seer 1.11 at 0000000168987100.
Waiting for Seer 1.11.
Eng02 (Seer 1.11) <- id name Seer 1.11
nChessVariant(1) = 1, id name.
Cancelling StartEngineTimerEvent() for Seer 1.11 at 0000000168987131.
Eng02 (Seer 1.11) <- id author Connor McMonigle
Eng02 (Seer 1.11) <- option name Clear Hash type button
Eng02 (Seer 1.11) <- option name Threads type spin default 1 min 1 max 512
Eng02 (Seer 1.11) <- option name Hash type spin default 128 min 1 max 65536
Eng02 (Seer 1.11) <- option name Weights type string default save.bin
Eng02 (Seer 1.11) <- uciok
nChessVariant(1) = 1, uciok.
SendToEng2Time 0000000168987147 : Eng02 (Seer 1.11) -> setoption name Threads value 1
.
SendToEng2Time 0000000168987162 : Eng02 (Seer 1.11) -> setoption name Hash value 128
.
SendToEng2Time 0000000168987162 : Eng02 (Seer 1.11) -> setoption name Weights value save.bin
.
Total Physical Memory = 4095 MB.
Total Available Memory = 1409 MB.
Hash + TbCache to allocate for Seer 1.11 = 144 MB.
SendToEng2Time 0000000168987162 : Eng02 (Seer 1.11) -> isready
...
Scheduling StartEngineTimerEvent() for Seer 1.11 at 0000000168988098.
Waiting for Seer 1.11.
Eng01 (Seer 1.11) <- id name Seer 1.11
nChessVariant(0) = 1, id name.
Cancelling StartEngineTimerEvent() for Seer 1.11 at 0000000168988129.
Eng01 (Seer 1.11) <- id author Connor McMonigle
Eng01 (Seer 1.11) <- option name Clear Hash type button
Eng01 (Seer 1.11) <- option name Threads type spin default 1 min 1 max 512
Eng01 (Seer 1.11) <- option name Hash type spin default 128 min 1 max 65536
Eng01 (Seer 1.11) <- option name Weights type string default save.bin
Eng01 (Seer 1.11) <- uciok
nChessVariant(0) = 1, uciok.
SendToEng1Time 0000000168988161 : Eng01 (Seer 1.11) -> setoption name Threads value 1
.
SendToEng1Time 0000000168988161 : Eng01 (Seer 1.11) -> setoption name Hash value 128
.
SendToEng1Time 0000000168988161 : Eng01 (Seer 1.11) -> setoption name Weights value save.bin

Code: Select all

Time Overstep Margin = 0 ms
Brain work may now start. Good Luck !
 ==========================================================================================
Eng01 (Seer 1.11) <- info depth 1 score cp 0 nodes 0 nps 0 time 2 pv
Eng01 (Seer 1.11) <- Pipe01Time 0000000168995415
Exact move time for Seer 1.11 
Eng01 (Seer 1.11) <- info depth 6 score cp 80 nodes 1403 nps 87687 time 15 pv g1f3 g8f6 c2c4 e7e6 b1c3 d7d5
Eng01 (Seer 1.11) <- info depth 7 score cp 76 nodes 3271 nps 102218 time 31 pv b1c3 d7d5 e2e3 g8f6 f1b5 c8d7 g1f3
Eng01 (Seer 1.11) <- info depth 8 score cp 80 nodes 7081 nps 110640 time 63 pv e2e4 g8f6 b1c3 a7a6 d2d4 b8c6 g1f3
Eng01 (Seer 1.11) <- info depth 9 score cp 83 nodes 8921 nps 114371 time 77 pv e2e4 c7c5 g1f3 e7e6 b1c3 b8c6 f1b5 g8f6 b5c6
Eng01 (Seer 1.11) <- info depth 10 score cp 75 nodes 21888 nps 116425 time 187 pv g1f3 g8f6 d2d4 d7d5 c1f4 b8c6 b1c3 e7e6 f4c7
Eng01 (Seer 1.11) <- info depth 11 score cp 63 nodes 34527 nps 115862 time 297 pv g1f3 g8f6 c2c4 c7c5 e2e3 b8c6 f1e2 e7e6 b1c3 f8e7 d2d4
Eng01 (Seer 1.11) <- info depth 12 score cp 59 nodes 52559 nps 115768 time 453 pv g1f3 g8f6 e2e3 b8c6 d2d4 e7e6 c2c4 f8e7 b1c3 d7d5 c4d5
Eng01 (Seer 1.11) <- info depth 13 score cp 45 nodes 135723 nps 116002 time 1169 pv g1f3 d7d5 e2e3 g8f6 c2c4 e7e6 b1c3 b8c6 f1e2 f8d6 d2d4 e8g8 e1g1
Eng01 (Seer 1.11) <- info depth 14 score cp 46 nodes 245106 nps 114642 time 2137 pv e2e4 e7e5 g1f3 b8c6 f1b5 g8f6 e1g1 d7d5 e4d5 f6d5 b5c6 b7c6
Eng01 (Seer 1.11) <- info depth 15 score cp 59 nodes 294411 nps 114378 time 2573 pv e2e4 e7e5 g1f3 b8c6 b1c3 g8f6 f1c4 f8b4 e1g1 d7d6 a2a4 b4c3
Eng01 (Seer 1.11) <- info depth 16 score cp 52 nodes 459548 nps 113300 time 4055 pv e2e4 e7e5 g1f3 b8c6 d2d4 e5d4 f3d4 g8f6 d4c6 d7c6 d1d8 e8d8 c1g5 h7h6 g5f6
Eng01 (Seer 1.11) <- bestmove e2e4
https://rwbc-chess.de

trollwatch:
Talkchess nowadays is a joke - it is full of trolls/idiots/people stuck in the pleistocene > 80% of the posts fall into this category...
RubiChess
Posts: 587
Joined: Fri Mar 30, 2018 7:20 am
Full name: Andreas Matthies

Re: Seer

Post by RubiChess »

I can confirm that Seer 1.1 doesn't work in ChessGUI here and even in Arena it fails to run or even configure. Only CuteChess seems to be tolerant with the missing default in Weights option.

Andreas
User avatar
Sylwy
Posts: 4468
Joined: Fri Apr 21, 2006 4:19 pm
Location: IASI - the historical capital of MOLDOVA
Full name: SilvianR

Re: Seer

Post by Sylwy »

RubiChess wrote: Sat Nov 14, 2020 4:04 pm I can confirm that Seer 1.1 doesn't work in ChessGUI here and even in Arena it fails to run or even configure.
Andreas
In Arena 3.5.1 GUI Seer 1.1 works perfectly !

Image

Image
RubiChess
Posts: 587
Joined: Fri Mar 30, 2018 7:20 am
Full name: Andreas Matthies

Re: Seer

Post by RubiChess »

Sylwy wrote: Sat Nov 14, 2020 4:23 pm
RubiChess wrote: Sat Nov 14, 2020 4:04 pm I can confirm that Seer 1.1 doesn't work in ChessGUI here and even in Arena it fails to run or even configure.
Andreas
In Arena 3.5.1 GUI Seer 1.1 works perfectly !
You are right. Autodetection didn't work for Seer and Arena had set it as a Winboard engine. After switching to UCI I was able to use it.
User avatar
Guenther
Posts: 4611
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: Seer

Post by Guenther »

Guenther wrote: Sat Nov 14, 2020 12:44 pm ...

When installing ChessGUI sends a stop after it has gathered all options
and Seer reacts with bestmove h1h1 and then it crashes after ChessGUI sends quit.

I could replicate this behaviour from Seer in cmd.

Code: Select all

0.000: - Starting server: seer111-64.exe
0.031: - Priority class set to normal
0.047: - Thread priority set to normal
0.047: > uci
1.233: < id name Seer 1.11
1.233: < id author Connor McMonigle
1.233: < option name Clear Hash type button
1.233: < option name Threads type spin default 1 min 1 max 512
1.233: < option name Hash type spin default 128 min 1 max 65536
1.233: < option name Weights type string default save.bin
1.233: < uciok
1.373: > stop
1.373: < bestmove h1h1
1.373: > quit
1.373: < 
1.373: < This application has requested the Runtime to terminate it in an unusual way.
1.373: < Please contact the application's support team for more information.
3.370: - Client closed
3.370: - Stopping server communication
8.377: - Closing log
I have added an issue now for Connor for the bug with crashing during installation in ChessGUI.
https://github.com/connormcmonigle/seer-nnue/issues/1
https://rwbc-chess.de

trollwatch:
Talkchess nowadays is a joke - it is full of trolls/idiots/people stuck in the pleistocene > 80% of the posts fall into this category...
connor_mcmonigle
Posts: 533
Joined: Sun Sep 06, 2020 4:40 am
Full name: Connor McMonigle

Re: Seer

Post by connor_mcmonigle »

Thanks for all the investigative work relating to Seer failing to work ChessGUI. I'll make a new release shortly which hopefully will resolve these issues and should probably add ~60 elo including my progress over the last week.

I simply hadn't considered that a GUI might send a stop command before issuing go. I guess I just didn't read the UCI specification carefully enough. My attempts to determine the origin of the incompatibility with Chess GUI were largely thwarted by ChessGUI crashing long before I even had a chance to specify a path to my engine...

I'll also change the behavior of the "go" command such that it doesn't need any additional parameters and defaults to doing a short search. Currently Seer crashes if a go command was specified without any limits.
connor_mcmonigle
Posts: 533
Joined: Sun Sep 06, 2020 4:40 am
Full name: Connor McMonigle

Re: Seer

Post by connor_mcmonigle »

I've released Seer 1.2 here: https://github.com/connormcmonigle/seer ... s/tag/v1.2
It should resolve existing UCI compatibility issues (and hopefully doesn't create any new issues). The improved network adds significant elo over the previous release.