Are you not willing to say that Komodo is trained with separate code? I would think you would say so, because that would set Komodo apart.dkappe wrote: ↑Mon Nov 02, 2020 9:33 pmFor night nurse, yes.AndrewGrant wrote: ↑Mon Nov 02, 2020 9:32 pmSo, it is the same training code then, except you've cut out the SF bits?dkappe wrote: ↑Mon Nov 02, 2020 9:28 pm I think Minic maybe swapped in it’s own qsearch, otherwise I’m not aware of any net author other than myself that has replaced the qsearch. I can’t speak to dragon, but I replaced the qsearch for training night nurse with a very simple one. You’re not looking for the best qsearch for playing, just quiet positions for training.
Komodo 14.1 Release and Dragon Announcement
Moderators: hgm, Rebel, chrisw
-
- Posts: 1756
- Joined: Tue Apr 19, 2016 6:08 am
- Location: U.S.A
- Full name: Andrew Grant
Re: Komodo 14.1 Release and Dragon Announcement
Last edited by AndrewGrant on Mon Nov 02, 2020 9:36 pm, edited 1 time in total.
#WeAreAllDraude #JusticeForDraude #RememberDraude #LeptirBigUltra
"Those who can't do, clone instead" - Eduard ( A real life friend, not this forum's Eduard )
"Those who can't do, clone instead" - Eduard ( A real life friend, not this forum's Eduard )
-
- Posts: 512
- Joined: Tue Sep 29, 2020 4:29 pm
- Location: Dublin, Ireland
- Full name: Madeleine Birchfield
Re: Komodo 14.1 Release and Dragon Announcement
NNUE is a much bigger framework than just king-piece or half king-piece; king-piece/half king-piece is only one NNUE architecture amongst many that are possible. There exist NNUE architectures (currently unused as far as I know) that do not use king-piece whatsoever for input features, such as mobility or piece count NNUE architectures.AndrewGrant wrote: ↑Mon Nov 02, 2020 9:21 pm Well, Halogen's original networks are trained in Pytorch. Halogen's current network is trained by a private NN/NNUE trainer I wrote; but Halogen is NOT using the NNUE King-Piece, Shogi Style, whatever you want to call it, framework. All Halogen networks are trained on game results I believe, not evaluations, so there is no SF involved at any step.
Stockfish developers themselves have recently been talking about moving away from half king-piece to a slightly different NNUE architecture and adding more input features like passed pawns to their network.
-
- Posts: 55
- Joined: Sun Feb 04, 2018 12:38 pm
- Location: UK
Re: Komodo 14.1 Release and Dragon Announcement
Great news but as Frank already mentioned Arena bundles Dragon by Bruno Lucas even if its not actively developed now, so can you call it something else please.
Author of the actively developed PSYCHO chess engine
-
- Posts: 512
- Joined: Tue Sep 29, 2020 4:29 pm
- Location: Dublin, Ireland
- Full name: Madeleine Birchfield
Re: Komodo 14.1 Release and Dragon Announcement
Komodo Dragon would be a good name.IanKennedy wrote: ↑Mon Nov 02, 2020 9:37 pm Great news but as Frank already mentioned Arena bundles Dragon by Bruno Lucas even if its not actively developed now, so can you call it something else please.
-
- Posts: 1631
- Joined: Tue Aug 21, 2018 7:52 pm
- Full name: Dietrich Kappe
Re: Komodo 14.1 Release and Dragon Announcement
Not for me to say.AndrewGrant wrote: ↑Mon Nov 02, 2020 9:35 pm
Are you not willing to say that Komodo is trained with separate code? I would think you would say so, because that would set Komodo apart.
Fat Titz by Stockfish, the engine with the bodaciously big net. Remember: size matters. If you want to learn more about this engine just google for "Fat Titz".
-
- Posts: 1756
- Joined: Tue Apr 19, 2016 6:08 am
- Location: U.S.A
- Full name: Andrew Grant
Re: Komodo 14.1 Release and Dragon Announcement
Odd. I'm surprised by such a statement. Will Mark or Larry say?dkappe wrote: ↑Mon Nov 02, 2020 9:40 pmNot for me to say.AndrewGrant wrote: ↑Mon Nov 02, 2020 9:35 pm
Are you not willing to say that Komodo is trained with separate code? I would think you would say so, because that would set Komodo apart.
Otherwise people will assume this is just SF's trainer, like most/all other NNUEs in play at the moment.
#WeAreAllDraude #JusticeForDraude #RememberDraude #LeptirBigUltra
"Those who can't do, clone instead" - Eduard ( A real life friend, not this forum's Eduard )
"Those who can't do, clone instead" - Eduard ( A real life friend, not this forum's Eduard )
-
- Posts: 512
- Joined: Tue Sep 29, 2020 4:29 pm
- Location: Dublin, Ireland
- Full name: Madeleine Birchfield
Re: Komodo 14.1 Release and Dragon Announcement
Regarding Seer, this is what Connor had to say about Seer's architecture vs Stockfish:AndrewGrant wrote: ↑Mon Nov 02, 2020 9:21 pm I am not familiar with what Seer is doing, although I have talked to the author a fair amount.
connor_mcmonigle wrote: ↑Mon Nov 02, 2020 9:24 pm The tensor product is a generalization of the notion of an outer product (read: in this context they mean the same thing). Consequently, the question is a bit meaningless. I've not seen halfkp features described by a tensor product before, but one could consider it a tensor product of piece and king features if they wanted to.
If we let vec(x) denote flattening x, a X b represent the outerproduct of a and b, B represent the board features in {0, 1}^12x8x8 (a boolean "tensor"), and let B[0, :, :] and B[1, :, :] (using slice notation) denote white and black king planes respectively, then Seer's input is (vec(vec(B[0, :, :]) X vec(B)), vec(vec(B[1, :, :]) X vec(B))). This differs from the input features currently used in SF networks which would be (vec(vec(B[0, :, :]) X vec(B[:2, :, :])), vec(vec(B[1, :, :]) X vec(B[:2, :, :]))).
Current Stockfish input features are referred to by the name "halfkp" (king X piece). AFAIK, some individuals experimenting with switching to use Seer style input for Stockfish networks have coined the name "halfka" (king X all). Also notable is that, in Stockfish, the (black king) X (piece) features are rotated (though the rotating is an accidental, possibly elo gaining, bug as mirroring was intended). Seer doesn't use mirroring/rotating as it uses separate affine transforms ("feature transformers") for the white and black halves of the halfka input.
Last edited by Madeleine Birchfield on Mon Nov 02, 2020 9:49 pm, edited 1 time in total.
-
- Posts: 1631
- Joined: Tue Aug 21, 2018 7:52 pm
- Full name: Dietrich Kappe
Re: Komodo 14.1 Release and Dragon Announcement
You like to push things to a crisis, don’t you? Let me just say here, as I have many times in the SF discord and elsewhere, that the sooner SF can move away from a Hand crafted c++ CPU-based trainer to a tensorflow or pytorch, gpu based trainer, the better. I see they’ve started taking some baby steps in that direction.AndrewGrant wrote: ↑Mon Nov 02, 2020 9:41 pmOdd. I'm surprised by such a statement. Will Mark or Larry say?dkappe wrote: ↑Mon Nov 02, 2020 9:40 pmNot for me to say.AndrewGrant wrote: ↑Mon Nov 02, 2020 9:35 pm
Are you not willing to say that Komodo is trained with separate code? I would think you would say so, because that would set Komodo apart.
Otherwise people will assume this is just SF's trainer, like most/all other NNUEs in play at the moment.
Fat Titz by Stockfish, the engine with the bodaciously big net. Remember: size matters. If you want to learn more about this engine just google for "Fat Titz".
-
- Posts: 1756
- Joined: Tue Apr 19, 2016 6:08 am
- Location: U.S.A
- Full name: Andrew Grant
Re: Komodo 14.1 Release and Dragon Announcement
I agree with you on that. Pytorch is problematic, but I had assumed you trained your networks with Tensorflow.
That was a wrong assumption for Bad Gyal.
#WeAreAllDraude #JusticeForDraude #RememberDraude #LeptirBigUltra
"Those who can't do, clone instead" - Eduard ( A real life friend, not this forum's Eduard )
"Those who can't do, clone instead" - Eduard ( A real life friend, not this forum's Eduard )
-
- Posts: 1631
- Joined: Tue Aug 21, 2018 7:52 pm
- Full name: Dietrich Kappe
Re: Komodo 14.1 Release and Dragon Announcement
Bad Gyal is a leela style resnet initially trained on tensorflow, now trained on a pytorch based system. It has nothing to do with NNUE.AndrewGrant wrote: ↑Mon Nov 02, 2020 9:56 pmI agree with you on that. Pytorch is problematic, but I had assumed you trained your networks with Tensorflow.
That was a wrong assumption for Bad Gyal.
Fat Titz by Stockfish, the engine with the bodaciously big net. Remember: size matters. If you want to learn more about this engine just google for "Fat Titz".