Search found 293 matches

by Raphexon
Tue Jul 21, 2020 9:20 am
Forum: Computer Chess Club: General Topics
Topic: Alpha Zero's 21. Bg5! revisited
Replies: 83
Views: 8497

Re: Alpha Zero's 21. Bg5! revisited

Still, from a points a view...on a modest CPU beats Lc0, being slightly trained (at least compared to Lc0). Have you tested Lc0 vs NNUE yet? My GPU is just a lousy r9 380 so I'm having a hard time trying to create a scenario that doesn't take too much time, is representative and achieves balanced r...
by Raphexon
Mon Jul 20, 2020 11:13 pm
Forum: Computer Chess Club: General Topics
Topic: Alpha Zero's 21. Bg5! revisited
Replies: 83
Views: 8497

Re: Alpha Zero's 21. Bg5! revisited

Sergio's net: https://www.comp.nus.edu.sg/~sergio-v/nnue/ With halfkp 256x binary from 7-13: Threads 1 Hash 16 mb No TB Didn't even need 25 million nodes to find it. Absolutely incredible. FEN: rn3r1k/pn1p1ppq/bpp4p/7P/4N1Q1/6P1/PP3PB1/R1B1R1K1 w - - 3 21 Sf-nnue-bmi2-256halfkp - Sergio: 1/1 00:00 ...
by Raphexon
Mon Jul 20, 2020 3:54 pm
Forum: Computer Chess Club: General Topics
Topic: Alpha Zero's 21. Bg5! revisited
Replies: 83
Views: 8497

Re: Alpha Zero's 21. Bg5! revisited

Sergio's net: https://www.comp.nus.edu.sg/~sergio-v/nnue/ With halfkp 256x binary from 7-13: Threads 1 Hash 16 mb No TB Didn't even need 25 million nodes to find it. Absolutely incredible. FEN: rn3r1k/pn1p1ppq/bpp4p/7P/4N1Q1/6P1/PP3PB1/R1B1R1K1 w - - 3 21 Sf-nnue-bmi2-256halfkp - Sergio: 1/1 00:00 ...
by Raphexon
Mon Jul 20, 2020 2:46 pm
Forum: Computer Chess Club: General Topics
Topic: Alpha Zero's 21. Bg5! revisited
Replies: 83
Views: 8497

Re: Alpha Zero's 21. Bg5! revisited

Sergio's net: https://www.comp.nus.edu.sg/~sergio-v/nnue/ With halfkp 256x binary from 7-13: Threads 1 Hash 16 mb No TB Didn't even need 25 million nodes to find it. Absolutely incredible. FEN: rn3r1k/pn1p1ppq/bpp4p/7P/4N1Q1/6P1/PP3PB1/R1B1R1K1 w - - 3 21 Sf-nnue-bmi2-256halfkp - Sergio: 1/1 00:00 4...
by Raphexon
Mon Jul 20, 2020 10:10 am
Forum: Computer Chess Club: General Topics
Topic: Can the sardine! NNUE clobbers SF.
Replies: 59
Views: 7148

Re: Can the sardine! NNUE clobbers SF.

NNUE has no contempt. Contempt is located inside eval, and since it doesn't use SF's eval it doesn't have contempt. Maybe it has some learned contempt from SF games, but I don't think contempt with depth 8 and depth 12 games is too gamechanging. Depth 8 only has a few % drawrate either way, and way ...
by Raphexon
Sun Jul 19, 2020 7:23 pm
Forum: Computer Chess Club: General Topics
Topic: Can the sardine! NNUE clobbers SF.
Replies: 59
Views: 7148

Re: Can the sardine! NNUE clobbers SF.

Now I have 2 NNUE nets in the cluster, the FiNN02 384x2-32-32 net is even less related to anything compared to the KG 256x2-32-32 net. They seem to be close in strength. They aren't even closely related between themselves, 2 NNUE nets. https://i.postimg.cc/Y9NYTkh9/SFNNUE3.jpg Isn't that awesome? T...
by Raphexon
Sun Jul 19, 2020 3:13 pm
Forum: Computer Chess Club: General Topics
Topic: The Engine Crackers (4)
Replies: 33
Views: 2189

Re: The Engine Crackers (4)

I have no choice but to move this into Honey. sf-nnue-avx2-256halfkp below was my first run , using a precompiled version stockfish-nn was my own compilation tailored for the Threadripper. The above analysis was using 60 cores and about 20% of my two 2060 Supers. Does Stockfish-nnue use your GPUs a...
by Raphexon
Fri Jul 17, 2020 11:08 pm
Forum: Computer Chess Club: Tournaments and Matches
Topic: FGRL - 16 Cores vs RTX 2070 rating list - 7. StockfishNNUE GK 200627
Replies: 1
Views: 676

Re: FGRL - 16 Cores vs RTX 2070 rating list - 7. StockfishNNUE GK 200627

The latest binaries actually don't need different slowmover settings anymore.
Default works perfectly now.

But thanks for testing.
by Raphexon
Thu Jul 16, 2020 6:11 pm
Forum: Computer Chess Club: General Topics
Topic: Lizard-NNUE Experiment NOT bad with NNUE Net Evaluation.........
Replies: 53
Views: 5137

Re: Lizard-NNUE Experiment NOT bad with NNUE Net Evaluation.........

Also training from Komodo games won't approximate Komodo's eval and it certainly won't approximate Komodo's search. NNs don't just copy. Even if a Houdart would reverse-engineered* Komodo's search and add an NNUE trained on Komodo games it wouldn't play like Komodo. Although it probably does play in...