It is currently Fri Feb 26, 2021 9:48 pm

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 14 posts ] 
Author Message
Offline
 Post subject: KataGo stares at the empty board for a long time
Post #1 Posted: Mon Feb 17, 2020 5:18 am 
Lives in gote

Posts: 430
Location: Adelaide, South Australia
Liked others: 173
Was liked: 224
Rank: Australian 2 dan
GD Posts: 200
Just to humour Maharani, I left my computer on while I was out for a few hours. Specifically, I gave it 90 minutes to stare at an empty board with 7.5 komi, and another 90 minutes with 7 points (whole number) komi. Here are the log files.

7.5 komi
2020-02-17 18:46:14+1030: GTP Engine starting...
2020-02-17 18:46:14+1030: KataGo v1.3.2
2020-02-17 18:46:14+1030: Using 6 CPU thread(s) for search
2020-02-17 18:46:14+1030: nnRandSeed0 = 18136206439097047930
2020-02-17 18:46:14+1030: After dedups: nnModelFile0 = /opt/katago-1.3.2/g170-b20c256x2-s1913382912-d435450331.txt.gz useFP16 auto useNHWC auto
2020-02-17 18:46:19+1030: Found OpenCL Platform 0: NVIDIA CUDA (NVIDIA Corporation) (OpenCL 1.2 CUDA 10.0.185)
2020-02-17 18:46:19+1030: Found 1 device(s) on platform 0 with type GPU or Accelerator
2020-02-17 18:46:19+1030: Found OpenCL Device 0: GeForce GTX 1070 (NVIDIA Corporation)
2020-02-17 18:46:19+1030: Using OpenCL Device 0: GeForce GTX 1070 (NVIDIA Corporation) OpenCL 1.2 CUDA
2020-02-17 18:46:19+1030: Loaded tuning parameters from: /home/alex/.katago/opencltuning/tune6_gpuGeForceGTX1070_x19_y19_c256_mv8.txt
2020-02-17 18:46:19+1030: Loaded neural net with nnXLen 19 nnYLen 19
2020-02-17 18:46:19+1030: OpenCL backend: Model version 8
2020-02-17 18:46:19+1030: Loaded model /opt/katago-1.3.2/g170-b20c256x2-s1913382912-d435450331.txt.gz
2020-02-17 18:46:19+1030: GTP ready, beginning main protocol loop
2020-02-17 18:46:19+1030: Controller: komi 7.5
2020-02-17 18:46:19+1030: =
2020-02-17 18:46:19+1030: Controller: genmove b
2020-02-17 20:16:19+1030: MoveNum: 0 HASH: 7BF12F3F24903F0C225CD6C55BA8BA1F
A B C D E F G H J K L M N O P Q R S T
19 . . . . . . . . . . . . . . . . . . .
18 . . . . . . . . . . . . . . . . . . .
17 . . . . . . . . . . . . . . . . . . .
16 . . . . . . . . . . . . . . . @ . . .
15 . . . . . . . . . . . . . . . . . . .
14 . . . . . . . . . . . . . . . . . . .
13 . . . . . . . . . . . . . . . . . . .
12 . . . . . . . . . . . . . . . . . . .
11 . . . . . . . . . . . . . . . . . . .
10 . . . . . . . . . . . . . . . . . . .
9 . . . . . . . . . . . . . . . . . . .
8 . . . . . . . . . . . . . . . . . . .
7 . . . . . . . . . . . . . . . . . . .
6 . . . . . . . . . . . . . . . . . . .
5 . . . . . . . . . . . . . . . . . . .
4 . . . . . . . . . . . . . . . . . . .
3 . . . . . . . . . . . . . . . . . . .
2 . . . . . . . . . . . . . . . . . . .
1 . . . . . . . . . . . . . . . . . . .

koPOSITIONALscoreAREAtaxNONEsui1komi7.5
Time taken: 5400.01
Root visits: 3218689
NN rows: 1959974
NN batches: 658103
NN avg batch size: 2.97822
PV: Q16 Q4 D17 C4 E3 D5 D15 R14 O17 H3 R6 O4 F4 H5 C2 F5 G4 G5 H4
Tree:
: T -14.60c W -14.29c S -0.32c ( -1.2 L -0.8) N 3218689 -- Q16 Q4 D17 C4 E3 D5 D15
---Black(^)---
Q16 : T -14.38c W -14.09c S -0.29c ( -1.2 L -0.8) LCB -14.39c P 5.74% WF 0.45% PSV 1820124 N 1820124 -- Q16 Q4 D17 C4 E3 D5 D15 R14
Q4 : T -14.66c W -14.34c S -0.32c ( -1.2 L -0.8) LCB -14.70c P 5.71% WF 0.41% PSV 144889 N 214462 -- Q4 D4 Q16 D16 F17 O3 C6 R6
D4 : T -14.69c W -14.36c S -0.32c ( -1.2 L -0.8) LCB -14.73c P 5.90% WF 0.41% PSV 139413 N 190214 -- D4 Q4 C16 Q16 F16 F3 C6 D3
D17 : T -14.91c W -14.56c S -0.35c ( -1.2 L -0.8) LCB -14.96c P 8.54% WF 0.39% PSV 120831 N 120927 -- D17 Q16 D4 Q4 D14 F3 C6 D3
C16 : T -14.91c W -14.56c S -0.35c ( -1.2 L -0.8) LCB -14.96c P 7.99% WF 0.39% PSV 113575 N 113677 -- C16 D4 Q16 Q4 F16 R14 O17 R16
Q3 : T -14.91c W -14.56c S -0.35c ( -1.2 L -0.8) LCB -14.96c P 7.69% WF 0.39% PSV 108482 N 108572 -- Q3 D4 Q16 D16 Q6 O17 R14 F17
D16 : T -14.75c W -14.42c S -0.33c ( -1.2 L -0.8) LCB -14.80c P 5.44% WF 0.40% PSV 107139 N 124793 -- D16 Q16 C4 Q4 F4 F17 C14 O3
C4 : T -14.97c W -14.61c S -0.36c ( -1.2 L -0.8) LCB -15.02c P 8.10% WF 0.39% PSV 103477 N 103547 -- C4 D16 Q4 Q16 F4 R6 O3 R14
R4 : T -15.02c W -14.65c S -0.36c ( -1.3 L -0.8) LCB -15.07c P 8.54% WF 0.39% PSV 101779 N 101843 -- R4 Q16 D4 D16 O4 C6 F3 C4
Q17 : T -14.97c W -14.61c S -0.36c ( -1.2 L -0.8) LCB -15.02c P 7.87% WF 0.39% PSV 100724 N 100793 -- Q17 D16 Q4 D4 Q14 O3 R6 Q3

2020-02-17 20:16:20+1030: = Q16
2020-02-17 20:16:21+1030: All cleaned up, quitting


7 points komi
2020-02-17 20:26:21+1030: GTP Engine starting...
2020-02-17 20:26:21+1030: KataGo v1.3.2
2020-02-17 20:26:21+1030: Using 6 CPU thread(s) for search
2020-02-17 20:26:21+1030: nnRandSeed0 = 2134096094977361051
2020-02-17 20:26:21+1030: After dedups: nnModelFile0 = /opt/katago-1.3.2/g170-b20c256x2-s1913382912-d435450331.txt.gz useFP16 auto useNHWC auto
2020-02-17 20:26:27+1030: Found OpenCL Platform 0: NVIDIA CUDA (NVIDIA Corporation) (OpenCL 1.2 CUDA 10.0.185)
2020-02-17 20:26:27+1030: Found 1 device(s) on platform 0 with type GPU or Accelerator
2020-02-17 20:26:27+1030: Found OpenCL Device 0: GeForce GTX 1070 (NVIDIA Corporation)
2020-02-17 20:26:27+1030: Using OpenCL Device 0: GeForce GTX 1070 (NVIDIA Corporation) OpenCL 1.2 CUDA
2020-02-17 20:26:27+1030: Loaded tuning parameters from: /home/alex/.katago/opencltuning/tune6_gpuGeForceGTX1070_x19_y19_c256_mv8.txt
2020-02-17 20:26:27+1030: Loaded neural net with nnXLen 19 nnYLen 19
2020-02-17 20:26:27+1030: OpenCL backend: Model version 8
2020-02-17 20:26:27+1030: Loaded model /opt/katago-1.3.2/g170-b20c256x2-s1913382912-d435450331.txt.gz
2020-02-17 20:26:27+1030: GTP ready, beginning main protocol loop
2020-02-17 20:26:27+1030: Controller: komi 7
2020-02-17 20:26:27+1030: =
2020-02-17 20:26:27+1030: Controller: genmove b
2020-02-17 21:56:27+1030: MoveNum: 0 HASH: 7BF12F3F24903F0C225CD6C55BA8BA1F
A B C D E F G H J K L M N O P Q R S T
19 . . . . . . . . . . . . . . . . . . .
18 . . . . . . . . . . . . . . . . . . .
17 . . . . . . . . . . . . . . . . . . .
16 . . . . . . . . . . . . . . . . . . .
15 . . . . . . . . . . . . . . . . . . .
14 . . . . . . . . . . . . . . . . . . .
13 . . . . . . . . . . . . . . . . . . .
12 . . . . . . . . . . . . . . . . . . .
11 . . . . . . . . . . . . . . . . . . .
10 . . . . . . . . . . . . . . . . . . .
9 . . . . . . . . . . . . . . . . . . .
8 . . . . . . . . . . . . . . . . . . .
7 . . . . . . . . . . . . . . . . . . .
6 . . . . . . . . . . . . . . . . . . .
5 . . . . . . . . . . . . . . . . . . .
4 . . . @ . . . . . . . . . . . . . . .
3 . . . . . . . . . . . . . . . . . . .
2 . . . . . . . . . . . . . . . . . . .
1 . . . . . . . . . . . . . . . . . . .

koPOSITIONALscoreAREAtaxNONEsui1komi7
Time taken: 5400.01
Root visits: 3294048
NN rows: 1952261
NN batches: 656018
NN avg batch size: 2.97593
PV: D4 Q16 Q3 D16 Q6 C6 F3 C14 R14 O16 F17 E17 F16 K17 F13 R15 Q14 Q10
Tree:
: T -2.38c W -2.34c S -0.03c ( -0.2 L -0.2) N 3294048 -- D4 Q16 Q3 D16 Q6 C6 F3
---Black(^)---
D4 : T -2.16c W -2.15c S -0.01c ( -0.2 L -0.2) LCB -2.18c P 5.74% WF 0.45% PSV 2164072 N 2164072 -- D4 Q16 Q3 D16 Q6 C6 F3 C14
Q16 : T -2.52c W -2.47c S -0.04c ( -0.2 L -0.2) LCB -2.56c P 5.53% WF 0.41% PSV 117996 N 142429 -- Q16 D16 D4 Q4 R6 C3 D3 C4
D17 : T -2.78c W -2.70c S -0.08c ( -0.3 L -0.2) LCB -2.83c P 8.39% WF 0.39% PSV 105777 N 106251 -- D17 Q16 D4 Q4 D14 F3 C6 D3
Q4 : T -2.56c W -2.51c S -0.05c ( -0.2 L -0.2) LCB -2.60c P 5.29% WF 0.40% PSV 101925 N 108573 -- Q4 D4 D16 Q16 O17 C17 D17 C16
D3 : T -2.79c W -2.71c S -0.08c ( -0.3 L -0.2) LCB -2.84c P 8.00% WF 0.39% PSV 99115 N 99507 -- D3 Q4 D16 Q16 D6 F17 C14 D17
R4 : T -2.82c W -2.74c S -0.08c ( -0.3 L -0.2) LCB -2.87c P 8.28% WF 0.39% PSV 97607 N 97863 -- R4 Q16 D4 D16 O4 C6 F3 C4
C4 : T -2.84c W -2.75c S -0.08c ( -0.3 L -0.2) LCB -2.89c P 8.31% WF 0.39% PSV 95884 N 96082 -- C4 D16 Q4 Q16 F4 R6 R8 R14
Q17 : T -2.85c W -2.77c S -0.08c ( -0.3 L -0.2) LCB -2.91c P 8.28% WF 0.38% PSV 93274 N 93411 -- Q17 D16 Q4 D4 Q14 O3 R6 Q3
R16 : T -2.83c W -2.75c S -0.08c ( -0.3 L -0.2) LCB -2.88c P 7.88% WF 0.39% PSV 91580 N 91785 -- R16 Q4 D16 D4 O16 C14 F17 C16
Q3 : T -2.88c W -2.79c S -0.09c ( -0.3 L -0.2) LCB -2.93c P 8.18% WF 0.38% PSV 88973 N 89030 -- Q3 D4 Q16 D16 Q6 O17 R14 Q17

2020-02-17 21:56:27+1030: = D4
2020-02-17 21:56:29+1030: All cleaned up, quitting


I'm not exactly sure what all the numbers mean, but there are some hints in this github issue. For something like:
Code:
Q16 : T -14.38c W -14.09c S  -0.29c ( -1.2 L  -0.8) LCB  -14.39c P  5.74% WF  0.45% PSV 1820124 N 1820124

I think T=-14.38c is "total utility: 14.38 cents favouring white", and I'm guessing that cents run from -100 to +100, so to convert it to a winrate it's 50-(14.38/2) = 42.81% for black. And either -0.3 or -0.2 would be the score median, I'm not sure which one. N=1820124 would be the number of playouts for this move, and P=5.74% is the policy network value.

So 3 million playouts isn't enough for KataGo to stop preferring 4-4 points as the first move. It's interesting to see that the policy values are actually higher for 3-4 than for 4-4, which must mean that 4-4 gets preferred entirely on the basis of the winrates being higher.


This post by xela was liked by 2 people: Bill Spight, Maharani
Top
 Profile  
 
Offline
 Post subject: Re: KataGo stares at the empty board for a long time
Post #2 Posted: Mon Feb 17, 2020 6:32 am 
Dies in gote

Posts: 61
Liked others: 12
Was liked: 18
Universal go server handle: Jæja
TLDR: It's an interesting experiment to allow for many playouts, but improvements to the network could theoretically result in radically different results and we'll never know for sure.

Perhaps KataGo will start preferring another opening move when it has played enough games against itself and further improved the (value and policy) neural networks?

If I understand correctly, the playouts are generating moves and their win-rates by alternating between a policy network, which generates points of interest, and a value network, which estimates the value of the board (the win rate). Both these networks could theoretically change a lot if the training procedure ever gets KataGo out of a possible local optimum. This is a place in the search space of network configurations where all solutions in the immediate surrounding are suboptimal, hence solutions similar to the current one are preferred. Think of it as standing on a hilltop, but because of fog, you're unable to see whether there are even higher hilltops around us.

It could happen that the policy network will prefer different points for opening, thus selecting different targets for deep analysis. Also, the value network could be updated in such a way that different patterns and therefore different openings are preferred.

It's actually impossible to know whether a different, more optimal solution can be achieved. Even more so, it's even impossible to know whether we're currently in a global optimum. All we can do is keep improving and see what happens. However, it could be that we're searching around the Mont Blanc and we're unable to see Mount Everest :scratch:


Last edited by Jæja on Mon Feb 17, 2020 8:10 am, edited 1 time in total.

This post by Jæja was liked by: Bill Spight
Top
 Profile  
 
Offline
 Post subject: Re: KataGo stares at the empty board for a long time
Post #3 Posted: Mon Feb 17, 2020 8:05 am 
Lives in gote

Posts: 641
Liked others: 105
Was liked: 762
Rank: maybe 2d
You guessed right with respect to the output. :salute: :)

Code:
<move> : T <total utility> W <winrate utility from -1 to +1> S <score utility from -something to +something> ( <selfplay score estimate> L <lead estimate>) LCB <lcb> P <policy> WF <slight weighting factor for move> PSV <value used to select move>  N <visit count>


This post by lightvector was liked by: Bill Spight
Top
 Profile  
 
Offline
 Post subject: Re: KataGo stares at the empty board for a long time
Post #4 Posted: Mon Feb 17, 2020 6:26 pm 
Lives in gote

Posts: 430
Location: Adelaide, South Australia
Liked others: 173
Was liked: 224
Rank: Australian 2 dan
GD Posts: 200
Thanks!

What's the difference between score estimate and lead estimate? If I had to take another guess :-) I'd guess that the first one is mean and the second is median?

PSV looks similar but not identical to visit count (sometimes but not always the same number). Is it some combination of visit count and utility?

Top
 Profile  
 
Offline
 Post subject: Re: KataGo stares at the empty board for a long time
Post #5 Posted: Mon Feb 17, 2020 6:43 pm 
Lives in gote

Posts: 641
Liked others: 105
Was liked: 762
Rank: maybe 2d
It's the same thing that I announced with the 1.3 release of KataGo.

ScoreSelfPlay is the estimated average score that would result from self-play, with self-play itself being affected by Kata not being entirely score-maximizing, sometimes taking risks or playing safe and giving up bits of points doing so.

Lead is the estimated average score adjustment that would be needed to make the winrate 50%. It's not the median of anything directly, it's trained be "how much do I need to adjust the komi to make myself (older versions of my self that generated my training data, plus a little bit of search) say 50%"? Since 1.3, this is the number that gets shown in GUIs.

Top
 Profile  
 
Offline
 Post subject: Re: KataGo stares at the empty board for a long time
Post #6 Posted: Tue Feb 18, 2020 12:16 am 
Lives with ko

Posts: 185
Liked others: 14
Was liked: 59
Rank: KGS 1k EGF 2k
KGS: Schachus12
One question about the lead: what does it mean, if it is a fractional number? Is it linearly interpolated between .5 of a whole number?

Top
 Profile  
 
Offline
 Post subject: Re: KataGo stares at the empty board for a long time
Post #7 Posted: Tue Feb 18, 2020 7:12 am 
Lives in gote

Posts: 641
Liked others: 105
Was liked: 762
Rank: maybe 2d
Yeah, sorta. There's some fiddly stuff that happens due to discreteness. Particularly with Chinese rules. I'm going off memory, so some of the below is perhaps not *quite* correct, since implementation was tricky.

Imagine you're playing black and that the neural net's best guess is that you're 40% to win on the board by 9 points, and 60% to win on the board by 7 points, and no other outcome is possible with bot-quality play. What is the "fair" komi?

Well, a komi of 7 would mean you get 70% equity (40% win + 60% draw), so that's not fair. A komi of 8 would mean you get 40% equity (40% win + 60% loss). So that's not fair either. The most fair blend would be if komi were 7 a third of the time, and komi were 8 two-thirds of the time, giving you 50% equity. So the neural net will be trained to try to say something like 7.666666... as fair in this case. In that case, if you're playing 7.5 komi, the neural net might say the lead is +0.16666666...

Yes that's a little weird, given that at 7.5 komi it would be saying the lead is +0.1666666, and your winrate would only be 40%. If komi were 8 or 8.5, your winrate would stay at 40%, but it would now be saying -0.33333333 or -0.833333333. This weirdness goes away in Japanese rules, since only in Chinese rules do you tend to have discreteness in chunks of 2 points.

Also, the neural net itself is a little noisy, so take all the above and add noise. And then MCTS will average across all these fractional values too, the same way it averages across winrates (since averaging is much cheaper than medianing) so there's a little bit of average-like behavior going on from the search itself.

So it's a bit messy. For most of the game, things are all smooth enough that you can interpret the number intuitively. Like, +0.7 means KataGo's opinion is that it's leading on average by 0.7 points, and if it goes down to +0.6, then KataGo's opinion changed downward by 0.1 point. That doesn't mean its opinion is correct, or that its opinion changed for a good reason - as Bill would say, the "margin of error" is certainly more than 0.1 points - and if the misunderstanding is major, such as a group on the board that it doesn't realize the status of due to a blind spot, it could be off by a ton more. But it would be accurate to think of its *opinion* as having changed slightly, even if just due to noise. Near the end of the game though, the tiny fractional differences are going to be basically averaging out the discreteness in game results and possible komi values, given KataGo's remaining uncertainty about the position.


This post by lightvector was liked by 5 people: Bill Spight, ez4u, Maharani, Waylon, xela
Top
 Profile  
 
Offline
 Post subject: Re: KataGo stares at the empty board for a long time
Post #8 Posted: Tue Feb 18, 2020 7:47 am 
Dies with sente

Posts: 81
Liked others: 19
Was liked: 46
lightvector wrote:
Well, a komi of 7 would mean you get 70% equity (40% win + 60% draw), so that's not fair. A komi of 8 would mean you get 40% equity (40% win + 60% loss). So that's not fair either. The most fair blend would be if komi were 7 a third of the time, and komi were 8 two-thirds of the time, giving you 50% equity.


I am a bit confused by this part. What does "equity" mean here? How is it computed? Also, does the estimated win rate just mean black's chance to win or the chance that either player may win? Couldn't the bot also estimate that both black and white have some chance of winning?

Top
 Profile  
 
Offline
 Post subject: Re: KataGo stares at the empty board for a long time
Post #9 Posted: Tue Feb 18, 2020 8:00 am 
Lives in gote

Posts: 641
Liked others: 105
Was liked: 762
Rank: maybe 2d
If you define a draw to be as good as 50% win and 50% of a loss, then 40% chance to win and 60% chance to draw:

40% + 0.5 * 60% = 70%.

Top
 Profile  
 
Offline
 Post subject: Re: KataGo stares at the empty board for a long time
Post #10 Posted: Sun Feb 23, 2020 1:05 pm 
Lives with ko

Posts: 130
Location: Pasadena, USA
Liked others: 55
Was liked: 8
Rank: OGS 11 kyu
OGS: Maharani
I've finally managed to get to enough playouts (1.2 million) within sixty minutes on ZBaduk to "fill in" this picture :) Every point on the third or fourth line of the board received at least eleven playouts. Only once this had happened did KataGo first consider points (i. e. give them more than ten playouts) that didn't have a third- or fourth-line coordinate (namely, the 5-5 points).

https://i.ibb.co/61C9cD0/Screen-Shot-20 ... -21-AM.png

Top
 Profile  
 
Offline
 Post subject: Re: KataGo stares at the empty board for a long time
Post #11 Posted: Sun Feb 23, 2020 1:47 pm 
Honinbo

Posts: 10660
Liked others: 3586
Was liked: 3337
Maharani wrote:
I've finally managed to get to enough playouts (1.2 million) within sixty minutes on ZBaduk to "fill in" this picture :) Every point on the third or fourth line of the board received at least eleven playouts. Only once this had happened did KataGo first consider points (i. e. give them more than ten playouts) that didn't have a third- or fourth-line coordinate (namely, the 5-5 points).

https://i.ibb.co/61C9cD0/Screen-Shot-20 ... -21-AM.png


Thanks. :)

Hmmmm. Should we consider the 3-4 pt. to have gotten 606k rollouts, and the 4-4 pt. to have gotten 567k rollouts?

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Top
 Profile  
 
Offline
 Post subject: Re: KataGo stares at the empty board for a long time
Post #12 Posted: Mon Feb 24, 2020 12:53 am 
Lives in gote

Posts: 430
Location: Adelaide, South Australia
Liked others: 173
Was liked: 224
Rank: Australian 2 dan
GD Posts: 200
Bill Spight wrote:
Should we consider the 3-4 pt. to have gotten 606k rollouts, and the 4-4 pt. to have gotten 567k rollouts?

Sorry, no. The rollouts in each corner are mostly duplicating the same information.

Top
 Profile  
 
Offline
 Post subject: Re: KataGo stares at the empty board for a long time
Post #13 Posted: Mon Feb 24, 2020 12:58 am 
Lives in gote

Posts: 430
Location: Adelaide, South Australia
Liked others: 173
Was liked: 224
Rank: Australian 2 dan
GD Posts: 200
Maharani wrote:
I've finally managed to get to enough playouts (1.2 million) within sixty minutes on ZBaduk to "fill in" this picture :) Every point on the third or fourth line of the board received at least eleven playouts. Only once this had happened did KataGo first consider points (i. e. give them more than ten playouts) that didn't have a third- or fourth-line coordinate (namely, the 5-5 points).

For this sort of thing, it's good to look at the network policy values, to give you a handle on how long it's going to take. Unfortunately I'm not sure how you do that with ZBaduk+KataGo. Room for future enhancements?

With Lizzie, the "show policy" button will show you policy values, but only for moves that have already got at least one playout. For LZ, you can see policy values for unexplored moves by running LZ from the command line and using the "heatmap" command. It would be nice to see this integrated into Lizzie some time.

I don't think KataGo yet has an equivalent to LZ's heatmap.

Top
 Profile  
 
Offline
 Post subject: Re: KataGo stares at the empty board for a long time
Post #14 Posted: Mon Feb 24, 2020 4:54 am 
Lives in gote

Posts: 641
Liked others: 105
Was liked: 762
Rank: maybe 2d
KataGo has it! See "kata-raw-nn" command documented at:

https://github.com/lightvector/KataGo/b ... ensions.md

Only in master branch for now (so needs custom compile), not part of a release yet. Will get included in next release, of course.


This post by lightvector was liked by 2 people: Maharani, xela
Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 14 posts ] 

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group