matejst wrote: ↑Tue Jun 30, 2020 2:49 am
Norm, I am not using bench: you know that it is too complicated for me. I was surprised when I noticed a surge in performance with the new binaries, two days ago -- on my old AMD A8, SF is achieves 550kn/s, the binaries from the 26th were ten times slower, about 53kn/s, and those of the 27th were achieving 155 kn/s. The new one are slower again. I also checked the number of positions after one mn, the depth, and it seems coherent with the nodes's count. Anyway, thank you for the popcount binaries -- it allowed me to finally try this new engine. It is similar to Winter, and recently Jonathan tried with two NN for the eval. I see a lot of intriguing, new possibilities arising.
Please try the new ones I just uploaded: 06-30-2020
These should be a lot faster.
matejst wrote: ↑Tue Jun 30, 2020 3:14 am
Thanks, Norman. They seem to be the fastest so far. Will there soon be a new NN?
I haven't really involved myself in training NN's, just run one once in awhile to test functionality.
Lot of experimenting at the moment.
It's still early in development, the codebase is changing every day...and getting better and better.
Now I rarely test in the Fritz GUI (switched to Banksia GUI), but maybe for someone this is important. Probably the developers should pay attention to this (if the problem is confirmed). In the Banksia GUI, all versions of Stockfish NN run without problems.
Now I rarely test in the Fritz GUI (switched to Banksia GUI), but maybe for someone this is important. Probably the developers should pay attention to this (if the problem is confirmed). In the Banksia GUI, all versions of Stockfish NN run without problems.
Hi Roman-
There are really only 2 things that have changed:
the newer 'popcnt' named compiles include support for SSE4.1 instruction set
and they load nn.bin upon 'ucinewgame' instead of the ''isready' command.
If the binary runs on your machine, and the GUI is UCI compliant (i.e issues 'ucinewgame')
there's no logical explanation for it to not work.
Some on sf-nnue discord have suggested Chessbase/Fritz are not 100% UCI compatible...
but that I know nothing about, I don't use them.
May I ask what happens when you double click on the executable?
kranium wrote: ↑Tue Jul 07, 2020 11:56 pm
If the binary runs on your machine, and the GUI is UCI compliant (i.e issues 'ucinewgame')
there's no logical explanation for it to not work.
...
May I ask what happens when you double click on the executable?
Hi kranium
The binary runs fine. Bench test passes. Most likely the problem is really in:
kranium wrote: ↑Tue Jul 07, 2020 11:56 pm
Some on sf-nnue discord have suggested Chessbase/Fritz are not 100% UCI compatible...
kranium wrote: ↑Tue Jul 07, 2020 11:56 pm
If the binary runs on your machine, and the GUI is UCI compliant (i.e issues 'ucinewgame')
there's no logical explanation for it to not work.
...
May I ask what happens when you double click on the executable?
Hi kranium
The binary runs fine. Bench test passes. Most likely the problem is really in:
kranium wrote: ↑Tue Jul 07, 2020 11:56 pm
Some on sf-nnue discord have suggested Chessbase/Fritz are not 100% UCI compatible...
Ok in command mode, crashes in Chessbase 15 if I try to use it for analysis - causes exception.
kranium wrote: ↑Tue Jul 07, 2020 11:56 pm
If the binary runs on your machine, and the GUI is UCI compliant (i.e issues 'ucinewgame')
there's no logical explanation for it to not work.
...
May I ask what happens when you double click on the executable?
Hi kranium
The binary runs fine. Bench test passes. Most likely the problem is really in:
kranium wrote: ↑Tue Jul 07, 2020 11:56 pm
Some on sf-nnue discord have suggested Chessbase/Fritz are not 100% UCI compatible...
Ok in command mode, crashes in Chessbase 15 if I try to use it for analysis - causes exception.
Hi John, Roman-
sf-nnue run (halfkp-256x2-32-32) x64-popc.exe
Per discussion on sf-nnue discord,
it's seems likely that Fritz/Chessbase can't handle long file names...
you might try shortening it