Search found 274 matches

by Raphexon
Tue Jul 28, 2020 9:03 pm
Forum: Computer Chess Club: General Topics
Topic: SF-NNUE going forward...
Replies: 79
Views: 6210

Re: SF-NNUE going forward...

People need several expensive and power-hungry GPUs to beat Stockfish What is interesting is that those people got the GPUs for Leela. They still have the GPUs. The GPUs didn't just disappear. And yet they're choosing to use their CPUs and NNUE nets because that gives them better performance (less ...
by Raphexon
Tue Jul 28, 2020 2:32 pm
Forum: Computer Chess Club: General Topics
Topic: cutechess gui build
Replies: 73
Views: 23660

Re: cutechess gui build

Gabor Szots wrote:
Tue Jul 28, 2020 11:56 am
There is a new version here: https://rwbc-chess.de/download.htm, by Günther Simon (Cli and GUI). Changes in the attached changes.txt.
Thank you very much for sharing.
by Raphexon
Tue Jul 28, 2020 10:00 am
Forum: Computer Chess Club: General Topics
Topic: New sf+nnue play-only compiles
Replies: 47
Views: 3414

Re: New sf+nnue play-only compiles

Hi all- I released new PO (play-only) compiles which seem to be significantly faster ...more than 10% on my system Intel® Core™ i9-9900K Processor Bench was run just once, not averaged after several runs, so these are just estimates: PO ------------------------- sf+nnue-po.270720.halfkp_256x2-32-32...
by Raphexon
Tue Jul 28, 2020 9:58 am
Forum: Computer Chess Club: General Topics
Topic: The next revolution in computer chess?
Replies: 74
Views: 5568

Re: The next revolution in computer chess?

... An educated guess is that for the 256 and 384 networks there is an N beyond which it no longer improves. So you think the 256 and 384 Nets are the most bigger in the viewpoint of usefulness? This is a practical experience or it has a basement in theory? Until incremental updates no longer fit i...
by Raphexon
Mon Jul 27, 2020 12:12 pm
Forum: Computer Chess Club: General Topics
Topic: SF-NNUE going forward...
Replies: 79
Views: 6210

Re: SF-NNUE going forward...

This was true only for the earliest versions of SFNNUE. "Speeded up by omitting the difference calculation for the pieces removed from the board" https://github.com/nodchip/Stockfish/commit/2523f72ff963b160dac5779bb9c7215de129113f On my current PC on 1 core: SF from start does roughly 1370 knps, nn...
by Raphexon
Mon Jul 27, 2020 11:30 am
Forum: Computer Chess Club: General Topics
Topic: SF-NNUE going forward...
Replies: 79
Views: 6210

Re: SF-NNUE going forward...

Tuning search is always fine, tuning position- learning code based on hash- learning wouldn't make code much more complicated, I think, even if I'm not good enough in proramming to prove so. Andscacs had a nice hash- storage before SugaR had so, no interest in position- learning based on selected h...
by Raphexon
Mon Jul 27, 2020 11:05 am
Forum: Computer Chess Club: General Topics
Topic: SF-NNUE going forward...
Replies: 79
Views: 6210

Re: SF-NNUE going forward...

Tuning search is always fine, tuning position- learning code based on hash- learning wouldn't make code much more complicated, I think, even if I'm not good enough in proramming to prove so. Andscacs had a nice hash- storage before SugaR had so, no interest in position- learning based on selected h...
by Raphexon
Fri Jul 24, 2020 10:24 pm
Forum: Computer Chess Club: General Topics
Topic: This is still pretty rough ... HONEY-NN release
Replies: 12
Views: 1208

Re: This is still pretty rough ... HONEY-NN release

What do you mean by "puzzle"? Study? Direct-mate? And why would a neural net be good at these? I would have thought that nets would be weaker than pure engines when the solution was purely tactical, as is the case for most compositions, particularly when the winning moves are often ipso facto impro...