Leela - Dance of the knights

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
brianr
Posts: 394
Joined: Thu Mar 09, 2006 2:01 pm

Re: Leela - Dance of the knights

Post by brianr » Fri May 22, 2020 10:44 pm

Or, just try running with 1 or 2 and look at the NPS

Glarean
Posts: 175
Joined: Sun Oct 05, 2008 11:04 am
Contact:

Re: Leela - Dance of the knights

Post by Glarean » Sat May 23, 2020 12:39 am

brianr wrote:
Fri May 22, 2020 10:44 pm
Or, just try running with 1 or 2 and look at the NPS
There isn't much difference:

Analysis by Lc0 0.25.1-t60-3010 (1 Thread)

1.d4 Sf6
Die Stellung ist ausgeglichen: = (0.12) Tiefe: 1/2 00:00:02 3N
1.Sf3 Sf6 2.d4 e6
Die Stellung ist ausgeglichen: = (0.09) Tiefe: 2/3 00:00:02 6N
Die Stellung ist ausgeglichen: = (0.10) Tiefe: 3/4 00:00:02 18N
1.d4 Sf6 2.c4 e6 3.Sf3 d5 4.Sc3 c5 5.cxd5 cxd4 6.Dxd4 exd5 7.Lg5 Le7 8.e3 0-0 9.Le2 Sc6 10.Dd3 h6 11.Lh4 Db6 12.0-0 Td8 13.Tfd1
Die Stellung ist ausgeglichen: = (0.12) Tiefe: 3/5 00:00:02 55N
Die Stellung ist ausgeglichen: = (0.11) Tiefe: 12/44 00:01:00 604kN

Analysis by Lc0 0.25.1-t60-3010 (2 Threads)

1.d4 Sf6
Die Stellung ist ausgeglichen: = (0.12) Tiefe: 1/2 00:00:02 3N
1.Sf3 Sf6 2.d4 e6
Die Stellung ist ausgeglichen: = (0.09) Tiefe: 2/3 00:00:02 6N
Die Stellung ist ausgeglichen: = (0.10) Tiefe: 2/4 00:00:02 19N
1.d4 Sf6 2.c4 e6 3.Sc3 Lb4 4.Sf3 0-0 5.e3 d5 6.Ld3 dxc4 7.Lxc4 c5 8.0-0 cxd4 9.exd4 b6 10.De2 Lb7 11.Lg5 Sbd7 12.Se5 Lxc3 13.bxc3
Die Stellung ist ausgeglichen: = (0.11) Tiefe: 2/4 00:00:02 28N
Die Stellung ist ausgeglichen: = (0.11) Tiefe: 12/43 00:01:00 649kN

brianr
Posts: 394
Joined: Thu Mar 09, 2006 2:01 pm

Re: Leela - Dance of the knights

Post by brianr » Sat May 23, 2020 1:31 am

The speed difference will also depend on the net.
3010 is a very large net, so it might be that 1 CPU is enough to keep the GPU busy.
The difference may well become larger with smaller net sizes.

In any case, nice to see Leela being used and tested more and more.

User avatar
cdani
Posts: 2107
Joined: Sat Jan 18, 2014 9:24 am
Location: Andorra
Contact:

Re: Leela - Dance of the knights

Post by cdani » Sat May 23, 2020 6:20 am

brianr wrote:
Sat May 23, 2020 1:31 am
In any case, nice to see Leela being used and tested more and more.
Of course is used! :-)

https://www.iccf.com/game?id=1146878

User avatar
Ovyron
Posts: 3950
Joined: Tue Jul 03, 2007 2:30 am

Re: Leela - Dance of the knights

Post by Ovyron » Sun May 24, 2020 4:17 am

Worth noting: SugarNN isn't an NN at all. It uses an A/B learning algorithm *called* NN ("Persistent Self-Learning") that has nothing to do with Neural Networks.

corres
Posts: 2592
Joined: Wed Nov 18, 2015 10:41 am
Location: hungary

Re: Leela - Dance of the knights

Post by corres » Mon May 25, 2020 7:45 pm

Ovyron wrote:
Sun May 24, 2020 4:17 am
Worth noting: SugarNN isn't an NN at all. It uses an A/B learning algorithm *called* NN ("Persistent Self-Learning") that has nothing to do with Neural Networks.
You are right.
Sugar "NN" is a false note.

corres
Posts: 2592
Joined: Wed Nov 18, 2015 10:41 am
Location: hungary

Re: Leela - Dance of the knights

Post by corres » Mon May 25, 2020 7:55 pm

Glarean wrote:
Sat May 23, 2020 12:39 am
brianr wrote:
Fri May 22, 2020 10:44 pm
Or, just try running with 1 or 2 and look at the NPS
There isn't much difference:

Analysis by Lc0 0.25.1-t60-3010 (1 Thread)

1.d4 Sf6
Die Stellung ist ausgeglichen: = (0.12) Tiefe: 1/2 00:00:02 3N
1.Sf3 Sf6 2.d4 e6
Die Stellung ist ausgeglichen: = (0.09) Tiefe: 2/3 00:00:02 6N
Die Stellung ist ausgeglichen: = (0.10) Tiefe: 3/4 00:00:02 18N
1.d4 Sf6 2.c4 e6 3.Sf3 d5 4.Sc3 c5 5.cxd5 cxd4 6.Dxd4 exd5 7.Lg5 Le7 8.e3 0-0 9.Le2 Sc6 10.Dd3 h6 11.Lh4 Db6 12.0-0 Td8 13.Tfd1
Die Stellung ist ausgeglichen: = (0.12) Tiefe: 3/5 00:00:02 55N
Die Stellung ist ausgeglichen: = (0.11) Tiefe: 12/44 00:01:00 604kN

Analysis by Lc0 0.25.1-t60-3010 (2 Threads)

1.d4 Sf6
Die Stellung ist ausgeglichen: = (0.12) Tiefe: 1/2 00:00:02 3N
1.Sf3 Sf6 2.d4 e6
Die Stellung ist ausgeglichen: = (0.09) Tiefe: 2/3 00:00:02 6N
Die Stellung ist ausgeglichen: = (0.10) Tiefe: 2/4 00:00:02 19N
1.d4 Sf6 2.c4 e6 3.Sc3 Lb4 4.Sf3 0-0 5.e3 d5 6.Ld3 dxc4 7.Lxc4 c5 8.0-0 cxd4 9.exd4 b6 10.De2 Lb7 11.Lg5 Sbd7 12.Se5 Lxc3 13.bxc3
Die Stellung ist ausgeglichen: = (0.11) Tiefe: 2/4 00:00:02 28N
Die Stellung ist ausgeglichen: = (0.11) Tiefe: 12/43 00:01:00 649kN
You would test it in more complicated positions and more smaller net, for e.g. 11248.

Post Reply