Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

PSQT .dat file has pawn structure patterns ? #14

Closed
tissatussa opened this issue Nov 23, 2022 · 5 comments
Closed

PSQT .dat file has pawn structure patterns ? #14

tissatussa opened this issue Nov 23, 2022 · 5 comments

Comments

@tissatussa
Copy link

i was wondering .. your pawn is the only engine i know which has a (.dat) file concerning Piece-SQuare-Tables .. and i don't remember previous pawn versions had such .. i guess it's part of a next version.

the .dat file is no plain text - it will contain those tables .. or does pawn have some NN now ? Or pawn structure patterns and such ?

you might not want to fully tell us your code ideas, i know ..

@ruicoelhopedro
Copy link
Owner

That file contains all the weights for the recent dynamic PSQ tables introduced in #12. Since they are quite a lot of values, I have packed them into that binary file (which has the .dat extension, but could have been something else). You can interpret them as some sort of NN, yes.

you might not want to fully tell us your code ideas, i know ..

If that was the case, this project wouldn't be open source ;)

@tissatussa
Copy link
Author

You can interpret them as some sort of NN, yes.

nice! are (only) self play games involved ?

@ruicoelhopedro
Copy link
Owner

nice! are (only) self play games involved ?

Yes, the training data is generated in self-play at a fixed shallow depth with some random moves to introduce diversity.

@AlexBrunetti
Copy link

You can interpret them as some sort of NN, yes.

This is an indexed table, as I can understand from the code; how can you call it a sort of NN?

Alex

@ruicoelhopedro
Copy link
Owner

In the sense that it contains the weights for a trained data-driven model used during evaluation. It's not an NN, but it is stored in a similar fashion as one would be.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants