It is currently Fri Feb 28, 2020 10:27 am

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 180 posts ]  Go to page Previous  1 ... 3, 4, 5, 6, 7, 8, 9  Next
Author Message
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #101 Posted: Wed Oct 30, 2019 3:14 am 
Dies with sente

Posts: 83
Liked others: 0
Was liked: 22
Rank: Tygem 7d
Marcel Grünauer wrote:
y_ich wrote:
Do you have good experiences of KataGo on PC?


I am running it on macOS 10.15 on a 2019 iMac; it works very well.


So you mean that KataGo works well for variable komi and better than Leela Zero with handicap stones.
Amazing.

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #102 Posted: Wed Oct 30, 2019 3:49 am 
Lives in gote

Posts: 628
Liked others: 264
Was liked: 316
y_ich wrote:
So you mean that KataGo works well for variable komi and better than Leela Zero with handicap stones.


I haven't really tested it with variable komi, but it works much better with handicap stones than Leela Zero. And it doesn't make ladder mistakes.

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #103 Posted: Fri Nov 01, 2019 12:44 pm 
Lives in sente
User avatar

Posts: 815
Liked others: 288
Was liked: 327
y_ich wrote:
wineandgolover san,

...

About No.3,
Could you give me a log with error?
I may help you or fix the problem if it is a bug.

Please understanding that posting about my flailing is a bit embarrassing.
Regarding not being able to see the json examples from your dropbox account, this is the error I get when I try to access "https://dl.dropboxusercontent.com/s/4ue9d86yaglengl/LeelaMaster_GX89_fp16.json" :
Attachment:
Screen Shot 2019-11-01 at 12.27.17 PM.png
Screen Shot 2019-11-01 at 12.27.17 PM.png [ 51.9 KiB | Viewed 1840 times ]


Furthermore, I assume I have to install the converter onto my computer, just as I did LZ and Lizzie, right? I assume that is how I will run this command:

Code:
./make.sh <board size> <URL of weight>


Below you can see the two commands I ran to try to do so (I also tried both commands with ".git" on the end.) I'm sure I made a simple and stupid mistake.

Code:
Last login: Fri Nov  1 12:01:25 on ttys001
Bradys-MacBook-Pro-2018:~ bradydaniels$ git clone -b next https://github.com/new3Rs/a_master_of_go
Cloning into 'a_master_of_go'...
fatal: Remote branch next not found in upstream origin
Bradys-MacBook-Pro-2018:~ bradydaniels$ git clone -b next https://github.com/new3Rs/a_master_of_go/LZ2mlmodel
Cloning into 'LZ2mlmodel'...
remote: Not Found
fatal: repository 'https://github.com/new3Rs/a_master_of_go/LZ2mlmodel/' not found
Bradys-MacBook-Pro-2018:~ bradydaniels$



This is why I asked for step-by-step instructions on how to convert networks, where the steps include installing the appropriate stuff from GitHub.

y_ich wrote:
You seem to want to get the very strongest AI at present.
Good, me too^^

But I want to point out that the difference of Elo rating between #249 and #242 is only 155 points, which is about 70% winrate.
(It sounds big but do you really want to choose best sparring partner among Tyson, Foreman, Ali, et al.? I am kidding.)
I will update the weight when the difference of rating become 200 point bigger.



My interest lies less in sparring with a better partner, and more with optimizing for the hardware I have.

I understand and agree with not updating too frequently, and it makes sense for the 40b because it's pretty recent. But given the new A13 processor, I think it is worthy to try a 20 block network, which I suspect will still be pretty fast. And once I can do that successfully, perhaps updating the 15 block is also worthwhile, since the old one is based on 226, six months old.

I'd be happy to do this myself if I can, and let you focus on more important things, like say, the KataGo conversion. Alas, I am stuck.

Thanks again for your software, and for putting up with me.

_________________
Want to see videos of low-dan mistakes and what to learn from them? Brady's Blunders

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #104 Posted: Fri Nov 01, 2019 6:43 pm 
Dies with sente

Posts: 83
Liked others: 0
Was liked: 22
Rank: Tygem 7d
wineandgolover san,

I am sorry for my poor description in README.md.

1. JSON in README.md is an example.
You should replace "https://dl.dropboxusercontent.com/s/4ue9d86yaglengl/LeelaMaster_GX89_fp16.json" to the URL for your own model which you will convert according to the instruction in README.md.
By the way, this means that you need your web server to put your model like Dropbox in the example.

2. Your understanding is right for make.sh, but you do not need to add "-b next" option when git clone.
Code:
git clone https://github.com/new3Rs/a_master_of_go

cd a_master_of_go/LZ2mlmodel

# make.sh assumes that pipenv has already been installed in your computer.
# If not, comment out the line below.
# brew install pipenv

./make.sh 19 <URL of weight>

# create and edit your JSON file.

# put JSON file and converted model in your web server.


I am developing these scripts on macOS and I am not sure that they work well on other platforms.
Please let me know your platform.

3. optimization
I am sorry that I was kidding.
You fun is my fun.

(Finding optimized hyper parameters(Cinit, softmax temperature, virtual loss, etc.) sounds very difficult.
AlphaZeo, Leela Zero, and ELF OpenGo projects studied it, experimented many games, and optimized their parameters. Then they got different conclusions.)

About 20b block networks, win‐loss records show that ELFv2 is still strongest, it is stronger than the latest 20b(15b) weight which learned 40b self-games.
But ELFv2 is slower than ELFv1, I think ELFv1 is the best 20b for "A Master of Go", so I keep it preinstalled.

Anyway, you can try conversions and enjoy to watch their games.
Please feel free to tell me your troubles in conversion.


This post by y_ich was liked by: wineandgolover
Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #105 Posted: Thu Dec 12, 2019 1:52 am 
Lives in gote

Posts: 628
Liked others: 264
Was liked: 316
I was happy to see a new version of "Master of Go" that includes KataGo as an in-app purchase. However, there seems to be a problem with buyin the upgrade.

Here is a screenshot of the dialog. It says "Invalid product IDs are detected."

Attachment:
KataGo-dialog.jpeg
KataGo-dialog.jpeg [ 72.95 KiB | Viewed 1472 times ]


Another question - does KataGo include the final ownership estimation and the scoreMean or do you still need to buy virtual coins to get a score estimation?

Thank you.

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #106 Posted: Thu Dec 12, 2019 3:23 pm 
Dies with sente

Posts: 83
Liked others: 0
Was liked: 22
Rank: Tygem 7d
Grünauer san,

I am sorry that unlocking KataGo is not available yet due to my mistake of procedure.
Please wait for a few days.

About coins, I have no idea about accuracy of KataGo's score estimation yet.
I have been using GnuGo estimation for a long time, at least it is quite good at grading.
So it is still there. You do not need to use them.
I hope that I may add whole game analysis on server some day. Coins will be used for it.

Thank you for your patience.

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #107 Posted: Fri Dec 13, 2019 12:48 pm 
Lives in sente
User avatar

Posts: 815
Liked others: 288
Was liked: 327
I am very excited that katago is coming to Master of Go.

Y_ich, I hope you are charging a single, up-front fee for this upgrade, rather than having ongoing fees. I think you should be rewarded for your efforts, and I’m happy to pay, but I personally feel disincentivized by apps that keep requesting money to utilize standard functions. For katago, the functions I’d expect to use the most are its built-in score estimation, variable komi, handicap capabilities, and ability to play well on various board sizes.

Out of interest, in your testing, is katago running pretty fast?

Thanks!

_________________
Want to see videos of low-dan mistakes and what to learn from them? Brady's Blunders

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #108 Posted: Fri Dec 13, 2019 3:59 pm 
Dies with sente

Posts: 83
Liked others: 0
Was liked: 22
Rank: Tygem 7d
wineandgolover san,

“Unlocking KataGo” is non-consumable. Once you buy it, you can use it forever with no charge.

KataGo on the app is not as fast as other weights.
I do not succeed to let it use Neural Engine yet.
Its performance is about 19 eps on the latest iPad Pro.
And currently the KataGo engine supports only 19x19.
I will extend to 9x9 and 13x13 later.

Now both the version 4.0.1 and KataGo item are available.

Enjoy!

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #109 Posted: Sat Dec 14, 2019 1:14 am 
Lives in gote

Posts: 628
Liked others: 264
Was liked: 316
I just bought the KataGo upgrade.

When tapping the score button in the lower right - the one with the scale icon - I assume that it overlays KataGo's own ownership estimation. This doesn't cost any coins.

But it also displays a pop-up dialog saying "keep touching". I did and then something else happened and a coin was deducted. You wrote that you use GNU Go estimation for paid scoring, but I greatly prefer KataGo's free and more accurate scoring.

I think it's not a good user interface decision to have a free function and a paid function on the the same button; a normal tap is free, a long tap costs money.

Would it be possible to disable the paid scoring when using KataGo?

Thank you.

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #110 Posted: Sat Dec 14, 2019 1:48 am 
Lives in gote

Posts: 628
Liked others: 264
Was liked: 316
Also, I have noticed strange behaviour in calculating the points when using KataGo.

I've entered a position using study mode. This produced screenshot 1 (below in the hidden section).

Then I exited the study with the red cross button, then entered it again. This produced screenshot 2.

I went back to the beginning, added some variation, then went back again and jumped to the end of the main line. This produced screenshot 3.

All three screenshots show the same board position.

In screenshot 1, it says +17.3 points for White in the win rate section.

In screenshot 2, it says +0.8 points for White.

In screenshot 3, it doesn't show any points.

The same position using KataGo with Lizzie shows "mean: -7.5 stdev: 22.2%".

Attachment:
File comment: Screenshot 1
Screenshot 1.png
Screenshot 1.png [ 417.06 KiB | Viewed 1375 times ]

Attachment:
File comment: Screenshot 2
Screenshot 2.png
Screenshot 2.png [ 413.68 KiB | Viewed 1375 times ]

Attachment:
File comment: Screenshot 3
Screenshot 3.png
Screenshot 3.png [ 412.98 KiB | Viewed 1375 times ]

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #111 Posted: Sat Dec 14, 2019 2:04 am 
Lives in gote

Posts: 628
Liked others: 264
Was liked: 316
(Separate post because I can't add more than three images in one post.)

Additionally, when stepping through the sequence, at one point it shows -11 points for Black (screenshot 4; Lizzie shows -7.8 point), but in the next move, +0.7 points for Black (screenshot 5; Lizzie shows. -8 points for Black).

Attachment:
File comment: Screenshot 4
Screenshot 4.png
Screenshot 4.png [ 407.7 KiB | Viewed 1406 times ]

Attachment:
File comment: Screenshot 5
Screenshot 5.png
Screenshot 5.png [ 415.32 KiB | Viewed 1406 times ]

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #112 Posted: Sat Dec 14, 2019 2:39 am 
Dies with sente

Posts: 83
Liked others: 0
Was liked: 22
Rank: Tygem 7d
Grünauer san,

I am glad to see that you seem enjoy the new feature.

I will consider the UI of the scale-icon button.

About pts. in the winrate bar, it is not a score.
It means how the latest move get or lose winrate percentage.
(It have been shown in previous versions.)

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #113 Posted: Sat Dec 14, 2019 9:42 am 
Lives in sente
User avatar

Posts: 815
Liked others: 288
Was liked: 327
Hi y_ich,

I bought the katago expansion because I want to support you and your efforts. I fully expected that the initial roll-out might feel a little beta-like, and am cool with that.

Right now, I think you have handicap and variable komi working, great job! Does it support katago score estimation, like that displayed on recent versions of Lizzie? This is one of the key strengths of katago over other bots.

I hope you make progress with the neural engine. This will make a huge speed difference. Does it look promising?

I bought the upgrade on my iPhone. How do I access it on my iPad (same apple ID)? It’s currently asking me to buy it again.

I expect I won’t use katago much until it supports scoring and gets faster. but I appreciate your efforts to make that happen. Thanks!

_________________
Want to see videos of low-dan mistakes and what to learn from them? Brady's Blunders

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #114 Posted: Sat Dec 14, 2019 9:55 am 
Lives in gote

Posts: 628
Liked others: 264
Was liked: 316
wineandgolover wrote:
Does it support katago score estimation, like that displayed on recent versions of Lizzie?


See one of my previous posts. You can tap the "scale" symbol that normally uses a coin to get an estimate, but with KataGo it displays the final ownership estimation for each intersection. Don't tap it too long though or it will use a coin.

I can't see the scoreMean, though.

wineandgolover wrote:
I bought the upgrade on my iPhone. How do I access it on my iPad (same apple ID)? It’s currently asking me to buy it again.


Tap the 'refresh' button in the top left of that dialog.

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #115 Posted: Sat Dec 14, 2019 2:26 pm 
Dies with sente

Posts: 83
Liked others: 0
Was liked: 22
Rank: Tygem 7d
wineandgolover san,

About two questions, Grünauer san's answers are correct. Thank you, Grünauer san!

Now the version 4.0.2 is available.
The 20b weight runs 3x faster than 4.0.1 thanks to Neural Engine.
It is still slower than Leela Zero weight. I will investigate whether further speed-up is possible.

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #116 Posted: Tue Dec 17, 2019 5:49 am 
Lives in gote

Posts: 628
Liked others: 264
Was liked: 316
y_ich-san,

I have a feature request. I would really like to be able to set up a position, then start playing from it against the AI.

So for example, it would be good to be able to tap the "Start Study" mode, play some moves, then open the current position in "Game" mode.

When you tap the red cross button in Study mode, it says "You can save this record via the share button in the main menu". If it also copied the SGF to the clipboard, you could then use this with the "Paste SGF for Game" button.

At the moment I have to use a separate SGF editor such as SmartGo Kifu, create the position there, copy it to the clipboard, then switch to "A Master of Go" and paste it there.

It would be relevant to start the game from the Study mode's currently displayed position; that way you could open a whole game in study mode, then go to the move that interests you and start playing from that position without having to delete later moves. This is different from "Paste SGF for Game" because there it always starts the game from the final position.

What do you think about this feature?

Thank you.

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #117 Posted: Wed Dec 18, 2019 2:56 am 
Dies with sente

Posts: 83
Liked others: 0
Was liked: 22
Rank: Tygem 7d
Grünauer san,

I will consider about your request.

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #118 Posted: Wed Dec 25, 2019 1:40 am 
Lives in gote

Posts: 628
Liked others: 264
Was liked: 316
y_ich-san,

I have a question about performance on the A12 vs A13 processors. That is because I am considering to buy an iPad Air 2019 or to wait for the iPad Pro 2020, even though the latter is probably going to be twice as expensive as the former.

I understand that both processors use the same eight-core Neural Engine, though the A13's one is faster and uses less energy. (The iPad Pro 2020 will probably have the A13X processor, but since there is no difference between the A12 and A12X processors in terms of Neural Engine I don't expect any difference between the A13 and A13X processors either.)

But "Master of Go" should have the same number of playouts on both the A12 and A13 processors, right?

Or does it use the GPU and CPU as well (which are different on A12 and A13)?

If the performance is the same then I might as well buy an iPad Air 2019.

Thank you.

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #119 Posted: Wed Dec 25, 2019 10:34 pm 
Beginner

Posts: 1
Liked others: 0
Was liked: 0
Rank: 8k
Hi,

Recently bought your app. Really great.

Question: Have you considered using one of these networks as an option? They are modeled on Leela zero but use smaller nets so it’s much faster:

https://github.com/leela-zero/leela-zer ... -552210675

Cheers!

Top
 Profile  
 
Offline
 Post subject: Re: A Go AI app for iOS with that strong neural network
Post #120 Posted: Thu Jan 02, 2020 1:17 am 
Dies with sente

Posts: 83
Liked others: 0
Was liked: 22
Rank: Tygem 7d
Mr25 san,

Happy new year.
And thank you for purchasing my app!

Yes, not the latest, but 15b weight is already available.
Please see https://github.com/new3Rs/a_master_of_go.
You can download weights in the setting menu which open by tapping the tao button at the left-top corner of main menu.
And the shortcut button to input the URL will help you.

Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 180 posts ]  Go to page Previous  1 ... 3, 4, 5, 6, 7, 8, 9  Next

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