I'm very curious how it translates to games.
Code: Select all
STS Size Net
--------------------------
2711 384x30 T40-1808
2686 384x30 T60-2-1538
2662 320x24 62000
2441 128x10 59670Search found 247 matches
- Mon Jan 20, 2020 11:34 am
- Forum: Computer Chess Club: General Topics
- Topic: STS rating v13.1 for Lc0 0.21.2 with nodes = 1
- Replies: 59
- Views: 16397
Re: STS rating v13.1 for Lc0 0.21.2 with nodes = 1
As first network, Sergio's latest 384x30-t40-1808.pb.gz scores over 2700 STS points (with nodes = 1).
I'm very curious how it translates to games.
I'm very curious how it translates to games.
- Tue Jan 14, 2020 10:41 am
- Forum: Computer Chess Club: General Topics
- Topic: Sugar NN for Mac
- Replies: 4
- Views: 1212
Re: Sugar NN for Mac
Looks like the author is aware of this issue (for Linux). But no solution yet.
- Sat Jan 11, 2020 3:04 pm
- Forum: Computer Chess Club: General Topics
- Topic: STS rating v13.1 for Lc0 0.21.2 with nodes = 1
- Replies: 59
- Views: 16397
Re: STS rating v13.1 for Lc0 0.21.2 with nodes = 1
Small update with some newer nets at different sizes. T60 rose slowly but steadily above 2650 points. Network 59533 took with 2378 points the lead over older 128x10 experiments. And Sergio's weights with T40 data are still on top. STS Size Net ------------------------------ 2677 384x30 T40-1573 2670...
- Sat Jan 11, 2020 11:10 am
- Forum: Computer Chess Club: General Topics
- Topic: Nibbler - Leela Chess Zero (Lc0) interface
- Replies: 40
- Views: 21580
Re: Nibbler - Leela Chess Zero (Lc0) interface
You've probably already noticed that Nibbler version 1.1.3 is available. It got an option to show WDL (win draw loss) stats in the infobox. This feature requires Lc0 0.23.x and recent Leela weights. Download links for Windows, Linux and macOS . The mac version includes Lc0 0.23.2 for OpenCL & BLAS.
- Fri Jan 10, 2020 3:13 pm
- Forum: Computer Chess Club: General Topics
- Topic: Lc0 OpenCL benchmark with 128x10 network
- Replies: 20
- Views: 6015
Re: Lc0 OpenCL benchmark with 128x10 network
I know this was for opencl benchmarks, but just for kicks I ran it with a 2060 RTX Super (cudnn-fp16) <snip> .. Benchmark final time 10.0043s calculating 110656 nodes per second. Thank you for ruining my day, Mike. :wink: How high is your OpenCL score? My new android smartphone LG G8s with a Qualco...
- Thu Oct 31, 2019 12:09 pm
- Forum: Computer Chess Club: General Topics
- Topic: STS rating v13.1 for Lc0 0.21.2 with nodes = 1
- Replies: 59
- Views: 16397
Re: STS rating v13.1 for Lc0 0.21.2 with nodes = 1
Policy head T60 surpassed T40 (net 42850 got 2581) at STS suite
Code: Select all
Net STS
----------
60030 1360
60065 1923
60100 2112
60150 2220
60200 2258
60300 2329
60400 2375
60500 2413
60660 2437
60780 2476 - last before LR drop
60880 2536
60960 2602- Fri Sep 06, 2019 4:10 pm
- Forum: Computer Chess Club: General Topics
- Topic: Slow Chess Blitz Classic
- Replies: 45
- Views: 10779
Re: Slow Chess Blitz Classic
Hi Jonathan, is there some hope for a mac version of Slow Chess Blitz Classic please? :) Using wine is atm no option, because your engine crashed immediately after sending uciok. $ wine slow64.exe uci id name SlowChess Blitz Classic 1.4b id author Jonathan Kreuzer option name Hash type spin default ...
- Fri Aug 30, 2019 7:08 am
- Forum: Computer Chess Club: General Topics
- Topic: Scorpio 3.0
- Replies: 60
- Views: 16580
Re: Scorpio 3.0
Played now several games between Scorpio and other engines within HIARCS Chess Explorer on macOS. All without any problems! There is a issue with the analysis mode (at least with HCE gui). After activating analysis, Scorpio (uci) shows it's thinking process on the position. Fine, but when going thro...
- Thu Aug 29, 2019 6:12 pm
- Forum: Computer Chess Club: General Topics
- Topic: Crystal, the new tactical monster
- Replies: 25
- Views: 8144
Re: Crystal, the new tactical monster
As Mike pointed out, there is no real anomalous behavior here. When the switch to Rh1 occurs, it is still a vast improvement from -6 to -1 (the search has not yet seen that Bxh7+ is winning). TBs usage can result in different search cutoffs at different plies & depths, which can add some variabilit...
- Thu Aug 29, 2019 6:02 pm
- Forum: Computer Chess Club: General Topics
- Topic: Crystal, the new tactical monster
- Replies: 25
- Views: 8144
Re: Crystal, the new tactical monster
same position with Honey-X5i with same setup - Honey does not see the solution and gets stick on the drawing Rh1 dep score nodes time (not shown: tbhits knps seldep) 42 +50.04 5.18G 7:56.10 Rh1 Qbxb1 Rxh7 Bh3+ Rxh3 Qxc1 Bh7+ Kh8 Qe1 gxf6 Bb1+ Kg8 Rh8+ Kg7 Rh7+ Kg8 41 +49.96 4.24G 6:28.62 Rh1 Qbxb1 ...