Search found 844 matches

by elcabesa
Sun Nov 22, 2020 2:58 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Place to find correct perft result from a fen position
Replies: 18
Views: 658

Re: Place to find correct perft result from a fen position

Ferdy wrote:
Sun Nov 22, 2020 2:49 pm
There are duplicates in perft.txt, 6752 is the number of unique positions.
yes, they come from a sampling from a pgn file. so there are duplicates, but it does the work :) it's a good set to check perft
by elcabesa
Sun Nov 22, 2020 1:35 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Place to find correct perft result from a fen position
Replies: 18
Views: 658

Re: Place to find correct perft result from a fen position

Tried to run the perft from elcabesa perft.txt, it failed on 27 out of 6752 positions on depth 1 and 2 only. Some positions where it failed. line: r3kb1r/p1p2ppp/2n4n/1p3bq1/P2pp1PP/1PN2P2/2PPP3/R1BQKBNR w KQkq -,26,1166,30119,1307169,34774744,1484852572, status: fail, depth: 2, nodes: 1141, data: ...
by elcabesa
Fri Nov 20, 2020 3:43 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Place to find correct perft result from a fen position
Replies: 18
Views: 658

Re: Place to find correct perft result from a fen position

Not questioning you, just found that some other I found online must be wrong :) Like this one: r1k1r2q/p1ppp1pp/8/8/8/8/P1PPP1PP/R1K1R2Q w KQkq - 0 1 It says it should be 23 at depth 1 but I can clearly count that it should be 21 for both black and white... Makes me wonder about the others I found ...
by elcabesa
Fri Nov 20, 2020 3:06 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Place to find correct perft result from a fen position
Replies: 18
Views: 658

Re: Place to find correct perft result from a fen position

eligolf wrote:
Fri Nov 20, 2020 3:02 pm
Thank you elca! Are these confirmed to be correct?
I hope so, Vajolet and Stockfish indipendently give the same result on those positions, so I suppose they are confirmed. But I haven't tried counting all the moves using a chessboard :)
by elcabesa
Fri Nov 20, 2020 3:04 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Place to find correct perft result from a fen position
Replies: 18
Views: 658

Re: Place to find correct perft result from a fen position

Which claims to be 2217 nodes at depth 6, but I get 2015 (for both black and white to move first). Is my code wrong, or are these random FENs I found online wrong? :) in this case the correct value is 2217. to debug your movegen and perft I suggest to implement "divide", you can find perft & divide...
by elcabesa
Fri Nov 20, 2020 2:53 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Place to find correct perft result from a fen position
Replies: 18
Views: 658

Re: Place to find correct perft result from a fen position

here you can find 6838 random position with perft values from depth 1 to 6 in a csv like format

https://github.com/elcabesa/vajolet/blo ... /perft.txt
by elcabesa
Thu Nov 19, 2020 6:36 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Pytorch NNUE training
Replies: 58
Views: 4450

Re: Pytorch NNUE training

thank you for your explanation and code let me recap what I have understood on factorizer so far: 1) you have 2 different network topologies, 1 one for training (TNNUE) and 1 for the engine (ENNUE) 2) you have a trick when saving the weight in the trainer so TNNUE and ENNUE have the same output for ...
by elcabesa
Tue Nov 17, 2020 12:05 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Pytorch NNUE training
Replies: 58
Views: 4450

Re: Pytorch NNUE training

Interesting! One of the experiments I had lined up was disabling the factorizer on the nodchip trainer and seeing how it did. But I’ll take your word for it :). I had already started implementing it, there are some really cool tricks the Shogi folks pulled off - zeroing the initial weights for the ...
by elcabesa
Mon Nov 16, 2020 6:05 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Pytorch NNUE training
Replies: 58
Views: 4450

Re: Pytorch NNUE training

Interesting! One of the experiments I had lined up was disabling the factorizer on the nodchip trainer and seeing how it did. But I’ll take your word for it :). I had already started implementing it, there are some really cool tricks the Shogi folks pulled off - zeroing the initial weights for the ...
by elcabesa
Sun Nov 15, 2020 8:37 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Pytorch NNUE training
Replies: 58
Views: 4450

Re: Pytorch NNUE training

I'm studying your code and Nodchip trainer. in nodchip trainer, the position from which the feature list is calculated doesn't seems to be the one stored in the gensfen bin file, but is the position resulting from a qsearch. I think this will help in resolving recapure and other not quiet positions ...