SPCC: Testrun of Fire 8.N Raptor finished

[pohl4711]

AB-testrun of Fire 8.N Raptor-net avx2 finished. The Raptor-net is not part of the official Fire 8.N-download, the Raptor-net was commited by the Fire-author 1 day after the release of Fire 8.N. Direct download-link to the Raptor-net on Fire-github:
https://github.com/FireFather/fire/raw/ ... x64/nn.bin

https://www.sp-cc.de

(Perhaps you have to clear your browsercache or reload the website)

[connor_mcmonigle]

The "Raptor" nn.bin network is not original either. It is just one of SV's earlier Stockfish networks: https://tests.stockfishchess.org/api/nn ... fd0df.nnue

I don't see why he gave it a new name. Norman didn't include this network in his official release so I can't comment as to whether he mislead you here. It's a little unfortunate that you spent your time testing it in any case.

[connor_mcmonigle]

Stefan seemed confused about the network, referring to it as the "raptor-net" his initial message. It seemed he thought it was a new network. I believe he was confused by this commit message, which, to his credit, is confusing: https://github.com/FireFather/fire/comm ... 5c4c00d0a2.

I didn't mean to accuse of anything and I think this was likely just a misunderstanding, but why the name "raptor"?

[mehmet123]

Fire 8N is now number 3 (except NN engines ) behind of Stockfish and Dragon. Ethereal NNUE is not an official version but it will be good to see this engine at rating lists.

[kranium]

The "Raptor" nn.bin network is not original either. It is just one of SV's earlier Stockfish networks: https://tests.stockfishchess.org/api/nn ... fd0df.nnue

I don't see why he gave it a new name. Norman didn't include this network in his official release so I can't comment as to whether he mislead you here. It's a little unfortunate that you spent your time testing it in any case.

for the record:
I added that Sergio Viera network to my release 5 days ago, on May 22, the same day I released the source code, albeit a bit afterward.
I named it Raptor during my testing and because it sounds sexy...
Details of this specific nn included in the Fire 8.N release download have been in the readme since day 1.
There has been no intention to mislead anyone

Here is a link to the commit:
https://github.com/FireFather/fire/comm ... f5077f13ce

Gerd and the guys at chessprogramming.org had no issue understanding the exact nature of the nn I included
https://www.chessprogramming.org/Fire
"Fire_8.N, released on May 22, 2021, features NNUE by utilizing a modified version of Daniel Shawul's Scorpio NNUE probing code along with a reinforcement learning net created by Sergio Vieri [16]."

Thanks much Stefan!

[connor_mcmonigle]

Fire 8N is now number 3 (except NN engines ) behind of Stockfish and Dragon. Ethereal NNUE is not an official version but it will be good to see this engine at rating lists.

I believe it is likely Nemorino, Rubichess and Igel would all jump by close to 100 elo in strength if tested with one of SV's networks as well. SV's results have been very difficult to replicate without training directly on evaluations produced by one of his networks.

...
I named it Raptor during my testing and because it sounds sexy...
...

I see

[Sylwy]

Fire 8N is now number 3 (except NN engines ) behind of Stockfish and Dragon. Ethereal NNUE is not an official version but it will be good to see this engine at rating lists.

........maybe no. 4 :
1.-excepting NN engines;
2.- behind FF2-Stockfish 13 & Komodo Dragon 1.0;
3.- excepting the originality of its NNUE (a SV one);
4.-............

[voffka]

Hello Connor,

[...]

I believe it is likely Nemorino, Rubichess and Igel would all jump by close to 100 elo in strength if tested with one of SV's networks as well. SV's results have been very difficult to replicate without training directly on evaluations produced by one of his networks.

[...]

Inspired by your post, I've just tested some of the Stockfish network - https://tests.stockfishchess.org/nns?page=2 which originated from SV I guess and inside Igel it gives a significant increase but not as big as 100 elo, around 40 elo at STC: http://chess.grantnet.us/test/11213/

Of course this network may perform better if scaling and search is optimized, perhaps some tweaks can bring it closer to 60 elo.

[kranium]

The "Raptor" nn.bin network is not original either. It is just one of SV's earlier Stockfish networks: https://tests.stockfishchess.org/api/nn ... fd0df.nnue

These networks are not 'Stockfish' networks...but I do understand your need to categorize them as such, making it seem as if anyone using them was taking from SF.

Some or many of Sergio's networks were certainly tested in Fishtest, but the vast majority were not successful, and did not become a default network.

To anyone interested, or to see if any particular SV network was tested by SF, or used elsewhere...all his networks are available here:
https://www.comp.nus.edu.sg/~sergio-v/nnue/

[AndrewGrant]

The "Raptor" nn.bin network is not original either. It is just one of SV's earlier Stockfish networks: https://tests.stockfishchess.org/api/nn ... fd0df.nnue

These networks are not 'Stockfish' networks...but I do understand your need to categorize them as such, making it seem as if anyone using them was taking from SF.

Some or many of Sergio's networks were certainly tested in Fishtest, but the vast majority (or all as far as I know including the one that I call Raptor) did not succeed, and did not become a default network.

To anyone interested, or to see if any particular SV network was tested by SF, or used elsewhere...all his networks are available here:
https://www.comp.nus.edu.sg/~sergio-v/nnue/

More than fair to call a network trained with Stockfish evals, from Stockfish games, in a trainer wrapped around Stockfish, a Stockfish Network