BrainLearn

[matejst]

Marco, I wish you all the best, and I hope we will chat here soon in better circumstances.

[amchess]

@amchess I rewrite it in more detail because maybe you haven't seen it

For information however:
in the infinity gui neither of the 2 engines works
an empty file is created in the installation folder of the gui and not the engine
The uci commands to activate the type of learning of Shashchess are not displayed in the options
Eman successfully creates the BIN file
both on Playchess and on Infinity
You are both programmers
I hope you will solve it

Eman does'n't have Q-learning.
Even in linux, for example, Arena doesn't display an uci option ("Petrosian").
Conversely, in windows, it works perfectly.
So, this is a gui bug, not the engine and this is never the first time.
For example, chessbase updates continually their guis also to solve this kind of problems.
Anyway, we can show you, after a match play in Fritz gui, experience files are created.
Sorry, but we can't do more than this.
Finally, as we wrote in the readme file, Q-learning is for selfplay: why to use it in playchess or infinitychess?

Andrea I refer to the standard mode
this problem was also present in version 11
I changed the code and deleted the new version drop down menu
have the possibility to compile a version with the standard default learning activated
in fritz gui it works well
while in the infinity GUI it creates an empty BIN file
also by installing the engine in the GUI installation folder

If InfinityChess gui can't even show the uci option, it's clear it's a gui problem, not the engine.
So, sorry, but we can't help you.

[Damir]

Hi Andrea

Eman is the only engine that uses its learnfile on Infinity Chess Server.

ShashChess and many other engines including Raubfish are not using its bin file. It might be a good idea to talk to Khalid Omar. Maybe he might be able to help you and Karl ( Raubfisch author) of what you are doing wrong, so learnfile is not used...

[amchess]

Hi Andrea

Eman is the only engine that uses its learnfile on Infinity Chess Server.

ShashChess and many other engines including Raubfish are not using its bin file. It might be a good idea to talk to Khalid Omar. Maybe he might be able to help you and Karl ( Raubfisch author) of what you are doing wrong, so learnfile is not used...

We published the code. So, any true programmer can propose what he wants.
We can't judge Eman engine and its learning because we don't have its code.
The only thing I noticed is the bin file is a lot bigger than our.
So, I suppose he uses an a lot different approach.
If he wants, he can help you, create a fork of our project and satisfy you: normally this is the philosophy of gpl code.
We don't trust in a gui clearly bugged if it never correctly displays some uci options.
So sorry, but we're very busy and don't want to waste our time.

[Damir]

Hi Andrea

Putting the blame on others for your engine not working the way it should, is not the right way to do. Instead you should openly acknowledge, that you have some limits regarding computer chess programming, and ask for help from some people who know more about the issue than you do, so they can help you fix the problem. it is no crime to ask for help.

[amchess]

Hi Andrea

Putting the blame on others for your engine not working the way it should, is not the right way to do. Instead you should openly acknowledge, that you have some limits regarding computer chess programming, and ask for help from some people who know more about the issue than you do, so they can help you fix the problem. it is no crime to ask for help.

I put the blame on nobody.
Everyone has limits about chess programming and programming in general, because nobody is perfect.
I asked for help and given help to other people very often in the past, even those self proclaming programmers, but only copy past experts.
I can name them one by one: ask them to write a simple c++ program and see if they're able.
In this case, we notice no bug because the problem is the gui, not the engine.
If you think the contrary, you have
1. to demonstrate it
2. to solve (because of 1) or ask someone other.
This is my last message about this theme, also because I'm not alone in the team and we're very busy.

[Zerbinati]

Marco, I wish you all the best, and I hope we will chat here soon in better circumstances.

Thanks Boban, better times will come.

[Ovyron]

Putting the blame on others for your engine not working the way it should

If the feature works as intended, but it doesn't work in the InfinityChess GUI, it's clear it's the GUI that is messing something up, if the GUI's programmer made engines work like in other GUIs, the learning would work in the InfinityChess GUI.

I bet you Eman's programmer just implemented a workaround so the feature would work, but unlike amchess, Khalid hides his sources and doesn't care about helping, just taking and taking, closing the source without giving anything away.

The best thing Khalid could do is opening the source, so just like he benefited by all the other open source projects he stole code from, people could benefit from his work. Give and take. But he's too egotistical for that, and prefers to break the GPL.

[matejst]

[...]
I bet you Eman's programmer just implemented a workaround so the feature would work, but unlike amchess, Khalid hides his sources and doesn't care about helping, just taking and taking, closing the source without giving anything away.
[...]

Uly,

I see you are very critical of Eman's, let's say, author lately. To tell the truth, I do not understand his behaviour, it makes absolutely no sense to me. Either you do your stuff and you assume fully, either you play by the rules. He is neither here nor there. Nonetheless, you wrote an excellent analysis of all the little things he has done to improve the usability and the strength of StockFish, and I hope a few programmers read it, since it could well be a fertile path to take. Andrea and his team, and Marco on his side, did good jobs already, let's hope they will continue with ShashChess and SugaR.

Anyway, that was not meant to be the topic of this post -- I do not use Eman and I am not interested in it -- but I rather wanted to ask you about the new hybrid engines, using a NN for eval and a more classical search, like SF NNUE and Winter. What's your take? Have you tried them?

[Ovyron]

I rather wanted to ask you about the new hybrid engines, using a NN for eval and a more classical search, like SF NNUE and Winter. What's your take? Have you tried them?

I haven't because they haven't proven yet to be better than classical evaluation. But their growth has been... let's say... scary. Imagine that they perform better than classical eval some day and the Stockfish team just replaces its eval for a network.

Then imagine if a similar concept appeared to create some sort of Network-driven search, somehow analyzing more important nodes and being able to not rely on big numbers to perform well.

Then all the years working on Stockfish's search and evaluation would be for nothing, in a world where everything becomes a black-box and nobody knows what the parameters are doing.

I don't like it because right now I can get to Stockfish's source code and change any of its evaluation parameters that I see fit, with a NN eval I don't think I can do anything but feeding it more games to get more accurate. Programming code basically becomes useless if better results can be achieved with just more training.

But this is something that has to be accepted as a wave of technology, if Stockfish's eval is like Newspapers and NN eval will be like the Internet, obviously it's the way of progress. But perhaps both branches will run alongside each other, and clever people will come up with new code for Stockfish's eval that will allow it to keep up with NNUE, just like Stockfish has managed to outpace Leela, unlike Winter that basically closed its eval, the only way to change it is to train it.