Sergio NN 40x512

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

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Hai
Posts: 548
Joined: Sun Aug 04, 2013 11:19 am

Re: Sergio NN 40x512

Post by Hai » Wed Feb 12, 2020 4:05 pm

corres wrote:
Wed Feb 12, 2020 10:54 am
Hai wrote:
Tue Feb 11, 2020 11:00 pm

After only 3 months of training it should reach 85 of 111 points.
The short Training time proves the new net strongly based on the earlier Lc0 training and because of this the weakpoints of the "new" net are the same as they are in Lc0 nets. So the new net in the reality an old net with some modification only. We can expect really improvement only from a fully new net I think.
I don't think so. You could also say Stockfish 11 in the reality is an old Engine with some modification only.
This big net has much more parameters to tune than the smaller nets. That's why it can be improved much longer before it reached his limit. That's also a reason why it will avoid most of the weakpoints we know from smaller nets.

I think 1 month, maybe + a half month, and not 3 months, will be enough to reach 85 of 111 points.
Looks like the 40x512 net needs only 2 days of training to solve one more of these extrem hard positions 8-) 8-) .

Hai
Posts: 548
Joined: Sun Aug 04, 2013 11:19 am

Re: Sergio NN 40x512

Post by Hai » Wed Feb 12, 2020 4:08 pm

ERET test suite.
40x512 (320)
64 GB RAM and a limit of 300 seconds per move: 88 of 111 points.
Again a big improvement :D .

Hai
Posts: 548
Joined: Sun Aug 04, 2013 11:19 am

Re: Sergio NN 40x512

Post by Hai » Wed Feb 12, 2020 5:52 pm

Hai wrote:
Wed Feb 12, 2020 4:08 pm
ERET test suite.
40x512 (320)
64 GB RAM and a limit of 300 seconds per move: 88 of 111 points.
Again a big improvement :D .
It is possible that the fourth 40x512 net will be better than 30x384 T40-2036, which only got 93 of 111 points :shock: :shock: :shock: .

corres
Posts: 3657
Joined: Wed Nov 18, 2015 10:41 am
Location: hungary

Re: Sergio NN 40x512

Post by corres » Wed Feb 12, 2020 6:49 pm

Hai wrote:
Wed Feb 12, 2020 4:05 pm
corres wrote:
Wed Feb 12, 2020 10:54 am
Hai wrote:
Tue Feb 11, 2020 11:00 pm

After only 3 months of training it should reach 85 of 111 points.
The short Training time proves the new net strongly based on the earlier Lc0 training and because of this the weakpoints of the "new" net are the same as they are in Lc0 nets. So the new net in the reality an old net with some modification only. We can expect really improvement only from a fully new net I think.
I don't think so. You could also say Stockfish 11 in the reality is an old Engine with some modification only.
This big net has much more parameters to tune than the smaller nets. That's why it can be improved much longer before it reached his limit. That's also a reason why it will avoid most of the weakpoints we know from smaller nets.

I think 1 month, maybe + a half month, and not 3 months, will be enough to reach 85 of 111 points.
Looks like the 40x512 net needs only 2 days of training to solve one more of these extrem hard positions 8-) 8-) .

Your simile is very primitive and very bad.
The power (~the knowlewdge) of a net is depend on very high measure of the self playing process. If somebody shorten the self playing process for.e.g. using other net the knowledge of the new net will not be much more higher than the knowledge of the base net. It is very obvious. The effect of those much more parameters will be really effective if they works during the self learning process also.

corres
Posts: 3657
Joined: Wed Nov 18, 2015 10:41 am
Location: hungary

Re: Sergio NN 40x512

Post by corres » Wed Feb 12, 2020 6:52 pm

corres wrote:
Wed Feb 12, 2020 6:49 pm
Hai wrote:
Wed Feb 12, 2020 4:05 pm
corres wrote:
Wed Feb 12, 2020 10:54 am
Hai wrote:
Tue Feb 11, 2020 11:00 pm

After only 3 months of training it should reach 85 of 111 points.
The short Training time proves the new net strongly based on the earlier Lc0 training and because of this the weakpoints of the "new" net are the same as they are in Lc0 nets. So the new net in the reality an old net with some modification only. We can expect really improvement only from a fully new net I think.
I don't think so. You could also say Stockfish 11 in the reality is an old Engine with some modification only.
This big net has much more parameters to tune than the smaller nets. That's why it can be improved much longer before it reached his limit. That's also a reason why it will avoid most of the weakpoints we know from smaller nets.

I think 1 month, maybe + a half month, and not 3 months, will be enough to reach 85 of 111 points.
Looks like the 40x512 net needs only 2 days of training to solve one more of these extrem hard positions 8-) 8-) .

Your simile is very primitive and very bad.
The power (~the knowlewdge) of a net is depend on very high measure of the self playing process. If somebody shorten the self learning process for.e.g. using other net the knowledge of the new net will not be much more higher than the knowledge of the base net. It is very obvious. The effect of those much more parameters will be really effective if they works during the self learning process also.

Hai
Posts: 548
Joined: Sun Aug 04, 2013 11:19 am

Re: Sergio NN 40x512

Post by Hai » Thu Feb 13, 2020 7:20 am

Hai wrote:
Wed Feb 12, 2020 8:09 am
Hai wrote:
Sat Jan 25, 2020 9:33 pm
LC0 Sergio 40x512 (175)
Hard Talkchess 2020 set

First result with 8 GB RAM and a limit of 15 seconds per move: 43 of 213 points = 20.1%.

But it is better in the ERET test suite.
LC0 Sergio 40x512 (320)
Hard Talkchess 2020 set
8 GB RAM and a limit of 15 seconds per move: 59 of 213 points = 27.7%.
From second (175) to third (320) net, from 43 to 59 solved positions means an improvement of +37.20%!! :lol: :lol: :lol:
We need some engine matches.

I think the fourth net will be ready in 2 weeks and much stronger 8-).
LC0 Sergio 40x512 (320)
Hard Talkchess 2020 set
64 GB RAM and a limit of 300 seconds per move: 82 of 213 points = 38.4%.

Hai
Posts: 548
Joined: Sun Aug 04, 2013 11:19 am

Re: Sergio NN 40x512

Post by Hai » Sat Mar 07, 2020 5:57 am

Do we have a new 40x512 net?

mmt
Posts: 343
Joined: Sun Aug 25, 2019 6:33 am
Full name: .

Re: Sergio NN 40x512

Post by mmt » Sat Mar 07, 2020 5:44 pm

I see a "320" here https://www.comp.nus.edu.sg/~sergio-v/test/512x40/. But in my testing 512x40 has been pretty disappointing so far. It will need a lot of training.

User avatar
Nordlandia
Posts: 2745
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: Sergio NN 40x512

Post by Nordlandia » Sat Mar 07, 2020 6:35 pm

New net is around the corner said Sergio. Under a weeks time.

At discord.

User avatar
Nordlandia
Posts: 2745
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: Sergio NN 40x512

Post by Nordlandia » Wed Mar 18, 2020 4:56 am

Sergio said he was at 1.5 out of 2.0 when i talked with him couple of days ago.

Second LR is 1-2 weeks away.

Post Reply