no reason for any more criticism.
On the other hand, I like more the closed source projects because I am often thinking ...
Dear programmers, please hold different good ideas secret, enough good ideas are public.
But different programmers like to claim proofs.
I think the most reason is that this group of programmers like to looking in the sources by the others.
That's allways the same. If sources are not available, the program is strong, it must be a clone.
NONSENSE!
Thank you Norman and good luck for the future of Fire.
thanks for the new version, including source code
I tested 8.0 privately under CEGT conditions. Does it make sense to re-test, i.e. is 8.1 stronger/different/weaker than 8.0?
Or can I use the games from 8.0 for our list?
I have no problem to re-test if you think it's better.
Wolfgang wrote: ↑Mon Apr 19, 2021 11:00 pm
Hi Norman,
thanks for the new version, including source code
I tested 8.0 privately under CEGT conditions. Does it make sense to re-test, i.e. is 8.1 stronger/different/weaker than 8.0?
Or can I use the games from 8.0 for our list?
I have no problem to re-test if you think it's better.
Thanks
Hi Wolf
There were minor code optimizations, so it may be a tiny bit faster...
of course such a small improvement would only really manifest itself in ultra-fast TCs.
I believe you will find little (if any) elo improvement in 8.1.
In normal play (at CEGT TCs) it should perform and test exactly like 8.0.
Looks to me as "original" as all the open source TOP 10 engines with their 1,000+ lines Stockfish like search functions.
Boring as hell.
So many "legit" SF authors have their "own" engines and competing in same events.
Should be 1 competition per 1 author per 1 engine.
JohnWoe wrote: ↑Tue Apr 20, 2021 10:34 am
Looks to me as "original" as all the open source TOP 10 engines with their 1,000+ lines Stockfish like search functions.
Boring as hell.
So many "legit" SF authors have their "own" engines and competing in same events.
Should be 1 competition per 1 author per 1 engine.
Yeah ! Happiness ! Each of the 10 authors must be absolutely sure that there is nothing original in the code. Otherwise it is, of course, a clone (if closed source). A sort of IT communism!
DanaSah 8.8 allows the use of a NNUE type neural network. It includes a network that has been created from another network created with CCRL 40/2 positions. More information in the readme.txt file. Although there are many changes, the version with the classic evaluation does not gain strength, about the same as version 8.3. The NNUE version with the dananet1 network win until 500 points, depending on the executable used (would gain up to about 600 with a stockfish net).
The network ccrl404_fn.nnue is created with all positions from the CCRL 40/2 list database. The games were converted to bin format and do not include evaluation and depth of the engines as the pgn database does not contain comments. For this reason the network is trained to learn by results. The network reminds me of the early lc0 networks where the game is very positional and less tactical. The network is not very strong, about 425 Elo points less than stockfish master, but it has possibilities to use it for new data and make it stronger.
pedrox wrote: ↑Tue Apr 20, 2021 9:19 pm
DanaSah 8.8 allows the use of a NNUE type neural network. It includes a network that has been created from another network created with CCRL 40/2 positions. More information in the readme.txt file. Although there are many changes, the version with the classic evaluation does not gain strength, about the same as version 8.3. The NNUE version with the dananet1 network win until 500 points, depending on the executable used (would gain up to about 600 with a stockfish net).
The network ccrl404_fn.nnue is created with all positions from the CCRL 40/2 list database. The games were converted to bin format and do not include evaluation and depth of the engines as the pgn database does not contain comments. For this reason the network is trained to learn by results. The network reminds me of the early lc0 networks where the game is very positional and less tactical. The network is not very strong, about 425 Elo points less than stockfish master, but it has possibilities to use it for new data and make it stronger.
pedrox wrote: ↑Tue Apr 20, 2021 9:19 pm
DanaSah 8.8 allows the use of a NNUE type neural network. It includes a network that has been created from another network created with CCRL 40/2 positions. More information in the readme.txt file. Although there are many changes, the version with the classic evaluation does not gain strength, about the same as version 8.3. The NNUE version with the dananet1 network win until 500 points, depending on the executable used (would gain up to about 600 with a stockfish net).
The network ccrl404_fn.nnue is created with all positions from the CCRL 40/2 list database. The games were converted to bin format and do not include evaluation and depth of the engines as the pgn database does not contain comments. For this reason the network is trained to learn by results. The network reminds me of the early lc0 networks where the game is very positional and less tactical. The network is not very strong, about 425 Elo points less than stockfish master, but it has possibilities to use it for new data and make it stronger.