TalkChess.com

Weak result is beating Leela on 4 cores now...

Just NN eval definitely isn't an opening book. But what about policy head? I definitely consider it book ish , Leela being very deterministic also supports it. NNUE just outputs a single eval. No policy. Also: https://github.com/dkappe/leela-chess-weights/wiki/Leela-is-an-opening-book-nonsense I wa...

Just NN eval definitely isn't an opening book.


But what about policy head?
I definitely consider it bookish, Leela being very deterministic also supports it.

zullil wrote: ↑

Tue Aug 04, 2020 4:45 pm

Leto wrote: ↑

Tue Aug 04, 2020 4:17 pm

Is that at LTC?

tc (20+0.2 )s

20+0.2 at 8 threads.

Taken from Hxim:

https://groups.google.com/g/fishcooking ... lHQQrbAwAJ

https://hxim.github.io/Stockfish-Evalua ... de/?p=nnue

Just upload a net (.bin files work too), input moves and watch how the values change.

At this time control (30 sec+ 0.5 ) all latest Stockfish NNUE versions easily passed +100 elo barrier against Stockfish 11 dev. At 1 min + 0.5 sec time control all latest Stockfish NNUE versions easily passed +90 elo barrier against Stockfish 11 dev. But in my tests the performance of latest Stockf...

ELO compression is the biggest factor probably. Take FastGM.de Difference between SF11 and Eth 12.25 is 253 elo at 60+0.6. At 60m+15s it's 170 elo. That doesn't mean Eth scales better than SF11, it's just less likely to blunder away a draw. You can replace Ethereal with most engines on that list. Th...

Rough maths suggests he has at least 1000 cores to do that. If we call several = 5 5 billion fens a day = 50000 fens per second, which if he samples at 25% implies 12500 games per second. One game every 0.08 ms. For massively fast bullet that needs 1000 cores at least, I think. 1000 cores would all...

Haven't read all of this thread so excuse me if it was already mentioned: tttak's fork has a lot of different architectures to try out. I'm interested in the "Gameply40x4" one which effectively holds 4 NNs in one file which get used sequentially switching every 40 plies until the 4th is loaded. Thi...

the_real_greco wrote: ↑

Wed Jul 29, 2020 2:12 pm

Efficiently-Updating Neural Network. It's backwards because of some wordplay in Japanese, if I remember correctly.

https://en.wikipedia.org/wiki/Nue