Page 3 of 6
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Tue Dec 24, 2019 8:11 am
by And
200 more games 9x9, sai with additional parameters:
100 games KataGo 10x128-s4588 - SAI (--lambda 0 --mu 0 --nrsymm), (gt 610, 5 sec/move) 55:45 (KataGo b 17:33 and KataGo w 38:12)
100 games KataGo 10x128-s4588 - SAI (--lambda 0 --mu 0 --nrsymm --puct 0), (gt 610, 5 sec/move) 84:16 (KataGo b 39:11 and KataGo w 45:5)
with these parameters SAI (--lambda 0 --mu 0 --nrsymm --puct 0) some games end very quickly:
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Tue Dec 24, 2019 1:39 pm
by And
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Wed Dec 25, 2019 5:09 am
by And
SAI 0.17.5
100 games KataGo 10x128-s4588 - SAI (--lambda 0 --mu 0 --nrsymm), (gt 610, 5 sec/move) 48:52 (KataGo b 10:40 and KataGo w 38:12)
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Thu Dec 26, 2019 3:52 am
by Vargo
Twenty 9x9 games
SAI 0.17.5 v. katago1.2
both : gogui-twogtp 1.5.1 , --noponder, komi 7.5, 7 sec/move (GTX 1080)
no duplicate game, no error, both program always agree on the score.
Katago wins 12-8 (60%)
networks :
SAI : 94619dea457de054503cec030269ce842c47055ba51e96db8fee841dfbaf05f9
katago : g104-b20c256-s447913472-d241840887
I'm not sure if these are the best weights available...
The first game, won by SAI (Black) :
The stats : (B+... always means SAI wins, even if SAi is W, because of the command -alternate)
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Thu Dec 26, 2019 4:53 am
by And
Vargo could you make some more games on your powerful graphics card, but SAI with parameters (--lambda 0 --mu 0 --nrsymm)?
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Thu Dec 26, 2019 5:26 am
by And
100 games 10 sec/move:
KataGo 10x128-s4588 - SAI (--lambda 0 --mu 0 --nrsymm), (gt 610) 51:49 (KataGo b 11:39 and KataGo w 40:10)
one of the games:
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Thu Dec 26, 2019 7:50 am
by Vargo
Second 20-game test (same parameters and hardware) :
SAI 0.17.5 v. katago1.2
Katago wins 13-7 (65%)
SAI : ... --lambda 0.5 -r 5 -t 6 --batchsize 6 --nrsymm , but not --mu 0, I'll add it next time.
stats :
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Mon Jan 06, 2020 6:05 am
by And
SAI #218 66e662fe (--lambda 0 --mu 0 --nrsymm)
http://sai.unich.it/networks/66e662fe4f ... 297ee7c.gz
against KataGo (10x128 s1108) 2:48 (5 sec/move, gt 610)
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Mon Jan 06, 2020 7:36 pm
by georgenan
SAI gives the only right answer to this
ALL zeros and its successor(leelaz\elfOpengo\minigo\ox\ on nextEngine;ttlEngine;alreadydoneEngine;komiEngine) cant give the right answer!!
Recently SAI weights strength goes beyond LZ092.
Yes, SAI really knows something now. Amazing!
You want to know the answer? Take SAI for a try！
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Wed Jan 08, 2020 12:58 pm
by And
georgenan wrote:SAI gives the only right answer to this
why is the only right answer? White can make any move and still win (maybe almost any
)
here
viewtopic.php?p=244851#p244851 really the only right answer
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Wed Jan 08, 2020 3:13 pm
by bugsti
And wrote:georgenan wrote:SAI gives the only right answer to this
why is the only right answer? White can make any move and still win (maybe almost any
)
here
viewtopic.php?p=244851#p244851 really the only right answer
Nope, you are probably forgetting the tiny differences between japanese and chinese rules
There is just one winning sequence for white under chinese rules.
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Wed Jan 08, 2020 3:40 pm
by Bill Spight
bugsti wrote:And wrote:georgenan wrote:SAI gives the only right answer to this
why is the only right answer? White can make any move and still win (maybe almost any
)
here
viewtopic.php?p=244851#p244851 really the only right answer
Nope, you are probably forgetting the tiny differences between japanese and chinese rules
There is just one winning sequence for white under chinese rules.
There are five winning moves for White under both Japanese and Chinese rules. (Edit: Corrected number of winning moves.) I haven't looked at the problem for a long time, but all problems in the book are White to play and win so that Japanese and Chinese rules agree. I.e., if White wins by 1 pt. of territory there will be an even number of stones on the board, the same number for each player.
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Thu Jan 09, 2020 3:22 pm
by And
Zenith Go 7 Analyze:
S14 - 52
Q10 - 51
H2 - 51
H16 - 49
G15 - 49
(fixed in SGF: Komi 0 and Chinese rules)
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Fri Jan 10, 2020 7:59 am
by Bill Spight
And wrote:Zenith Go 7 Analyze:
S14 - 52
Q10 - 51
H2 - 51
H16 - 49
G15 - 49
(fixed in SGF: Komi 0 and Chinese rules)
The five winning plays are S-14, Q-16, Q-07, Q-10, and R-09.
BTW, what is SAI's mainline solution?
Re: New networks Leela Zero, a Sensible Artificial Intellige
Posted: Fri Jan 10, 2020 10:10 am
by And
SAI 0.17.5 (Sabaki, 10000 Playouts, komi 0)
#222 afe36b9d
R9 -> 8280 (V: 53.90%)
O2 -> 431 (V: 50.05%)
H2 -> 236 (V: 51.91%)
S14 -> 193 (V: 49.90%)
Q16 -> 104 (V: 45.57%)
A7 -> 101 (V: 48.08%)
A17 -> 95 (V: 46.22%)
C19 -> 94 (V: 46.42%)
Q10 -> 85 (V: 52.01%)
#223 2cc322d9
S14 -> 8423 (V: 54.28%)
O2 -> 369 (V: 50.57%)
Q7 -> 217 (V: 49.62%)
Q10 -> 212 (V: 50.41%)
A17 -> 127 (V: 45.19%)
#224 dd403c0b
S14 -> 7354 (V: 54.74%)
H2 -> 1628 (V: 51.89%)
O2 -> 241 (V: 48.18%)
Q16 -> 169 (V: 46.91%)
C19 -> 114 (V: 45.18%)
#225 ef877be3
H2 -> 8419 (V: 54.16%)
O2 -> 409 (V: 49.71%)
S14 -> 355 (V: 51.01%)
Q16 -> 174 (V: 48.24%)
A17 -> 104 (V: 42.99%)
#230 6ebb71b9
S14 -> 6576 (V: 51.31%)
Q7 -> 2199 (V: 50.75%)
O2 -> 170 (V: 44.59%)
B8 -> 115 (V: 44.15%)
A7 -> 113 (V: 45.65%)