New J.A Compiles

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

Moderators: hgm, Rebel, chrisw

peter
Posts: 3206
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: New J.A Compiles

Post by peter »

Jim Ablett wrote: Tue May 28, 2024 3:01 pm Matefish 3.2 64 ja bmi2

Clang profiled build.

https://drive.proton.me/urls/QWCTHSHAX0#Kv1BFYx5oI4j
Wow, what a tempo!
Thank you
:!:
Peter.
schahmatist
Posts: 9
Joined: Sat Mar 12, 2022 3:52 pm
Full name: Dmitriy Fisch

Re: New J.A Compiles

Post by schahmatist »

Hi. Would it be possible, maybe, to compile Allie engine ?
https://github.com/manyoso/allie

Thank you!
User avatar
Jim Ablett
Posts: 1533
Joined: Fri Jul 14, 2006 7:56 am
Location: London, England
Full name: Jim Ablett

Re: New J.A Compiles

Post by Jim Ablett »

schahmatist wrote: Tue May 28, 2024 5:23 pm Hi. Would it be possible, maybe, to compile Allie engine ?
https://github.com/manyoso/allie

Thank you!
Image
Allie v0.7 by Adam Treat
https://github.com/manyoso/allie

You will need an Nvidia cuda equipped graphics card to run this.

Allie v0.7 64 ja

Windows build

https://drive.proton.me/urls/YT8TMW3650#o5YoQAY1BHWK

Jim.
Albert Silver
Posts: 3026
Joined: Wed Mar 08, 2006 9:57 pm
Location: Rio de Janeiro, Brazil

Re: New J.A Compiles

Post by Albert Silver »

Jim Ablett wrote: Tue May 14, 2024 7:54 pm
I am starting a new thread for any new engine compiles that I do, they will be posted here
instead of thread 'New engine releases & news H1 2024' which should be for new releases and updates
from original authors.


Image
Rodent NN by Fabien Letouzey, Thomas Gaksch, Jerry Donald Watson
Rebel NNUE evaluation added by Chris Whittington and Ed Schroder
In reality the Rodent net is actually 300 elo stronger when Pawel will implement the net into his own (Rodent) source code.
Not a new release but new compiles.

Rodent NN 64 ja

Windows & Linux profiled builds

https://drive.proton.me/urls/WYZ3P8J5R8#NgdZadRgzAda

Jim.
First of all, thanks for the great work. I got a bit lost in the descriptions. The last phrase said it is 300 Elo when implemented in the Rodent source code. Is that this build, or another one that is forthcoming? What is Letouzey's role here? The father of Fruit/Toga? Or something more recent?

Albert
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."
User avatar
Jim Ablett
Posts: 1533
Joined: Fri Jul 14, 2006 7:56 am
Location: London, England
Full name: Jim Ablett

Re: New J.A Compiles

Post by Jim Ablett »

Albert Silver wrote: Wed May 29, 2024 3:53 pm
Jim Ablett wrote: Tue May 14, 2024 7:54 pm
I am starting a new thread for any new engine compiles that I do, they will be posted here
instead of thread 'New engine releases & news H1 2024' which should be for new releases and updates
from original authors.


Image
Rodent NN by Fabien Letouzey, Thomas Gaksch, Jerry Donald Watson
Rebel NNUE evaluation added by Chris Whittington and Ed Schroder
In reality the Rodent net is actually 300 elo stronger when Pawel will implement the net into his own (Rodent) source code.
Not a new release but new compiles.

Rodent NN 64 ja

Windows & Linux profiled builds

https://drive.proton.me/urls/WYZ3P8J5R8#NgdZadRgzAda

Jim.
First of all, thanks for the great work. I got a bit lost in the descriptions. The last phrase said it is 300 Elo when implemented in the Rodent source code. Is that this build, or another one that is forthcoming? What is Letouzey's role here? The father of Fruit/Toga? Or something more recent?

Albert
Hi Albert,

This version is Fruit/Toga using Rodent eval+nnue. The version using original Rodent src + nnue which refers to 300 elo is this >

Rodent NNUE v1.0 by Pawel Koziol & Bernhard C. Maertz

https://drive.proton.me/urls/G26RXR5AMG#tGKkFzvJwtku

Jim.
Albert Silver
Posts: 3026
Joined: Wed Mar 08, 2006 9:57 pm
Location: Rio de Janeiro, Brazil

Re: New J.A Compiles

Post by Albert Silver »

Jim Ablett wrote: Wed May 29, 2024 4:41 pm
Albert Silver wrote: Wed May 29, 2024 3:53 pm
Jim Ablett wrote: Tue May 14, 2024 7:54 pm
I am starting a new thread for any new engine compiles that I do, they will be posted here
instead of thread 'New engine releases & news H1 2024' which should be for new releases and updates
from original authors.


Image
Rodent NN by Fabien Letouzey, Thomas Gaksch, Jerry Donald Watson
Rebel NNUE evaluation added by Chris Whittington and Ed Schroder
In reality the Rodent net is actually 300 elo stronger when Pawel will implement the net into his own (Rodent) source code.
Not a new release but new compiles.

Rodent NN 64 ja

Windows & Linux profiled builds

https://drive.proton.me/urls/WYZ3P8J5R8#NgdZadRgzAda

Jim.
First of all, thanks for the great work. I got a bit lost in the descriptions. The last phrase said it is 300 Elo when implemented in the Rodent source code. Is that this build, or another one that is forthcoming? What is Letouzey's role here? The father of Fruit/Toga? Or something more recent?

Albert
Hi Albert,

This version is Fruit/Toga using Rodent eval+nnue. The version using original Rodent src + nnue which refers to 300 elo is this >

Rodent NNUE v1.0 by Pawel Koziol & Bernhard C. Maertz

https://drive.proton.me/urls/G26RXR5AMG#tGKkFzvJwtku

Jim.
Ah fantastic, thanks!
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."
schahmatist
Posts: 9
Joined: Sat Mar 12, 2022 3:52 pm
Full name: Dmitriy Fisch

Re: New J.A Compiles

Post by schahmatist »

Jim Ablett wrote: Wed May 29, 2024 2:17 pm
schahmatist wrote: Tue May 28, 2024 5:23 pm Hi. Would it be possible, maybe, to compile Allie engine ?
https://github.com/manyoso/allie

Thank you!
Image
Allie v0.7 by Adam Treat
https://github.com/manyoso/allie

You will need an Nvidia cuda equipped graphics card to run this.

Allie v0.7 64 ja

Windows build

https://drive.proton.me/urls/YT8TMW3650#o5YoQAY1BHWK

Jim.
Thank you so much! Those executables is impossible to find! Alas, it says Wacatac.B!ml virus is there, cannot even unzip. Is it a common issue?
schahmatist
Posts: 9
Joined: Sat Mar 12, 2022 3:52 pm
Full name: Dmitriy Fisch

Re: New J.A Compiles

Post by schahmatist »

I added exception and was able to start the program. Still issues, looks like it just does random moves. Tried attached weights, as well as some of the Lc0 compatible weights still the same. I do have Windows 11 and Geforce RTX Nvidia


UPD: tested with the older laptop (Windows 10) - and it worked fine there. But not on windows 11 - only random moves there. Maybe it is not compatible with Windows 11...?

Code: Select all

id name Allie v0.7     (a0fb846)
id author Allie Chess Authors
option name CpuctBase type string default 15000
option name CpuctF type string default 2.817
option name CpuctInit type string default 2.1
option name DebugLog type check default false
option name DebugInfo type check default false
option name EarlyExitFactor type string default 0.72
option name FeaturesOff type string default none
option name Cache type spin default 5000000 min 100000 max 999999999
option name MaxBatchSize type spin default 272 min 0 max 65536
option name MoveOverhead type spin default 300 min 0 max 5000
option name GPUCores type spin default 1 min 0 max 256
option name OpeningTimeFactor type string default 2.15
option name Ponder type check default false
option name SyzygyPath type string default
option name TryPlayoutLimit type spin default 136 min 1 max 1000
option name UCI_Chess960 type check default false
option name UseFP16 type check default false
option name WeightsFile type string default C:/CHESS/ENGINES/2nd League/Allie/11258-120x10-se.pb.gz
uciok
go depth 2
info depth 1 seldepth 1 nodes 22 nps 2750 score cp 0 time 8 hashfull 0 tbhits 0 pv h2h3
info depth 1 seldepth 2 nodes 162 nps 12462 score cp 0 time 13 hashfull 0 tbhits 0 pv a2a4 h7h6
info depth 2 seldepth 3 nodes 993 nps 52263 score cp 0 time 19 hashfull 0 tbhits 0 pv g1h3 h7h6 a2a3
bestmove g1h3 ponder h7h6
Last edited by schahmatist on Wed May 29, 2024 7:42 pm, edited 2 times in total.
User avatar
AdminX
Posts: 6363
Joined: Mon Mar 13, 2006 2:34 pm
Location: Acworth, GA

Re: New J.A Compiles

Post by AdminX »

schahmatist wrote: Wed May 29, 2024 7:00 pm
Jim Ablett wrote: Wed May 29, 2024 2:17 pm
schahmatist wrote: Tue May 28, 2024 5:23 pm Hi. Would it be possible, maybe, to compile Allie engine ?
https://github.com/manyoso/allie

Thank you!
Image
Allie v0.7 by Adam Treat
https://github.com/manyoso/allie

You will need an Nvidia cuda equipped graphics card to run this.

Allie v0.7 64 ja

Windows build

https://drive.proton.me/urls/YT8TMW3650#o5YoQAY1BHWK

Jim.
Thank you so much! Those executables is impossible to find! Alas, it says Wacatac.B!ml virus is there, cannot even unzip. Is it a common issue?
I used 7-Zip to extract the files with no issues, give that a try.
"Good decisions come from experience, and experience comes from bad decisions."
__________________________________________________________________
Ted Summers
User avatar
Jim Ablett
Posts: 1533
Joined: Fri Jul 14, 2006 7:56 am
Location: London, England
Full name: Jim Ablett

Re: New J.A Compiles

Post by Jim Ablett »

schahmatist wrote: Wed May 29, 2024 7:25 pm I added exception and was able to start the program. Still issues, looks like it just does random moves. Tried attached weights, as well as some of the Lc0 compatible weights still the same. I do have Windows 11 and Geforce RTX Nvidia


UPD: tested with the older laptop (Windows 10) - and it worked fine there. But not on windows 11 - only random moves there. Maybe it is not compatible with Windows 11...?

Code: Select all

id name Allie v0.7     (a0fb846)
id author Allie Chess Authors
option name CpuctBase type string default 15000
option name CpuctF type string default 2.817
option name CpuctInit type string default 2.1
option name DebugLog type check default false
option name DebugInfo type check default false
option name EarlyExitFactor type string default 0.72
option name FeaturesOff type string default none
option name Cache type spin default 5000000 min 100000 max 999999999
option name MaxBatchSize type spin default 272 min 0 max 65536
option name MoveOverhead type spin default 300 min 0 max 5000
option name GPUCores type spin default 1 min 0 max 256
option name OpeningTimeFactor type string default 2.15
option name Ponder type check default false
option name SyzygyPath type string default
option name TryPlayoutLimit type spin default 136 min 1 max 1000
option name UCI_Chess960 type check default false
option name UseFP16 type check default false
option name WeightsFile type string default C:/CHESS/ENGINES/2nd League/Allie/11258-120x10-se.pb.gz
uciok
go depth 2
info depth 1 seldepth 1 nodes 22 nps 2750 score cp 0 time 8 hashfull 0 tbhits 0 pv h2h3
info depth 1 seldepth 2 nodes 162 nps 12462 score cp 0 time 13 hashfull 0 tbhits 0 pv a2a4 h7h6
info depth 2 seldepth 3 nodes 993 nps 52263 score cp 0 time 19 hashfull 0 tbhits 0 pv g1h3 h7h6 a2a3
bestmove g1h3 ponder h7h6
Does the laptop have an older nvidia graphics ?
You may need to replace the cudaNN dlls with ones for newer RTX cards.

Jim.