KaTrain: AI analysis/playing/teaching tool based on KataGo

Why?
I just need analysis for current move.

I opened 40 block model on weak small laptop. Even 1 visit for every of 200 moves of game is long time total. I review fuseki only - I need STOP button - so it don’t eats battery energy while thinking about moves I don’t care about.

Update:
Oh, there is delete node button - but “just stop thinking” button would be more elegant.

Nice added features, thank you.

Did Katrain open the menu before if you switched the window with Alt + Tab? It does now, anyway. I use it often, for example if I replay a game from a Youtube video. Using Alt as a shortcut isn’t a good idea I think.

Interestingly, the menu also appears when I switch with Win+Tab (left Win), although Win does not seem to be used in Katrain.

Why are you using the 40 block network? Basically only top-end machines can use that effectively.
Less blocks means each visit is less accurate, but significantly faster, leading to better results faster. Keep in mind kata usually requires around ten thousand visits to get trustworthy results.

1 Like

I like neural networks - its not soulless brute-force or monte-carlo tree search
Its truly made artificial intuition
I can’t beat 1 visit 40 block even with some stones handicap - so its more than enough
I guess patterns from such game may be more useful - they are less random than when ai uses many playouts - they are ignoring what normal ai takes in consideration - so its more easy to replicate to human

Any PC takes forever to get the big network up and running. And then the UI becomes unresponsive. In an equal CPU time match the smaller network tends to lose to a bigger network but that should only matter in a computer go tournament. So for an interactive user experience the smaller networks are much better. Even 15b is too strong for most users. A 5d teacher should be enough for a kuy player, no?

It used to be shift, but that interfered with other things. I should probably switch it to keyup instead of keydown.

I think, Alt or Shift would both be ok if you only react on key up and if no other key is pressed. This is also how the menu in Firefox works, for example.

V1.7.1 released with distributed training models and a whole bunch of bug fixes.

1.7.2 released with minor fixes, a new theme, tutorial, and an all-in-one OSX package.
By the way, if anyone is interested in creating themes, I am happy to help package them. I know it’s still a bit technical.

Oh man! Just discovered this. So excited to try it out tomorrow. Sounds like such a great idea. Bravo!

1 Like

v1.8 was just released with support for helping distributed training, and a feature for generating uneven positions to play from and practice your mid- and endgame skills.

5 Likes

v1.9 released, with a little competition on theming here

1 Like

v1.11 released

4 Likes

I’m late to the party; decided to start reviewing my games. I’ve seen Dwyrin using KaTrain, decided to give it a whirl…fully awesome. and intuitive as. Turns out there’s plenty of low-hanging fruit in my reviews.

Regarding the interface, in the AI suggested move bubbles, what is the lower (second line) number?
Does a higher number imply greater complexity, greater required reading depth, a lower number simplicity?

This example made me curious (though I’m not asking regarding this specific board position)…

Black has made numerous poor life decisions to even be in this position, but AI still calls it fairly even. I literally played the first move I saw (dotted kosumi Bottom Left) because it was shapey and cornery. Now, I notice that the playout number(?) for that move (507) is considerably higher than for the other options.

I assume it is number of visits (a.k.a. playouts).
It represents the amount of reading that the AI spent on follow-up variations. The associated score of a move with a high number of visits is more reliable than the scores of alternative moves with lower visit numbers (on which the AI spent a smaller amount of reading).
The reason why the AI spent more visits on the dotted move may be similar as yours (that move looked nice to the AI). But it may be a bit slow, as the AI seems to be changing its mind towards the jump (with the blue circle).

2 Likes

I suspect KaTrain has some feature where it automatically puts more visits on the move that was actually played in the game. (I don’t have much experience with KaTrain myself and I couldn’t find a setting for this - but just by playing through a game it seems like this must be the case. I hope someone more knowledgeable can step in and clarify how this works.)

This feature is useful since (as @gennan said) the more visits a move has, the more accurate its score will be, and you will often be interested in comparing your actual move with alternatives the AI is suggesting.

But it’s a bit confusing for someone used to other interfaces, where a high number of visits always means that the AI has decided “on its own” that the move looked good enough to be worth exploring to that extent. In KaTrain, even if you played a really bad move, it will get a high number of visits just because you played it.

So to accurately compare your move to other moves, look at the score of other moves that also got a decent number of visits. For instance in this case, there’s a number of moves to choose from that are about 1 point better than the actual played move - the AI thinks your move is okay, but not one of the strongest moves. And the higher number of visits doesn’t say anything about the move itself, it’s just the AI spending extra time thinking about it to make sure that it gives you an accurate number to compare with the other moves.

3 Likes

Cheers, reliability of estimate makes more sense than sequence complexity.

funnily enough, KaTrain has called me on this move now several times…I can do the small knight slap thing way more than I thought it turns out. Anyone who hasn’t typically reviewed their games, I highly recommend…I think I got more out of a few honest self-reviews than 100 games flinging stones.

4 Likes

The lower number is visits by default, though you can configure which two numbers to show.

[quote=“antonTobi, post:39, topic:30258”]
I suspect KaTrain has some feature where it automatically puts more visits on the move that was actually played in the game[/quote]

KaTrain always shows the most reliable number available for the score estimate for a move. Since the move was played, it has the analysis one move deeper, which has the full number of visits for an analysis.

1 Like

Please give me how to use it. And how to choose to use katago without using the GPU?
thank you.

Installation instructions for windows and download links are here: Releases · sanderland/katrain · GitHub

It runs fine without GPU. (At least the somewhat older version I have installed.)

2 Likes