jhonnold / berserk

UCI Chess Engine written in C
GNU General Public License v3.0
198 stars 30 forks source link

Can I train a separate eval network? #557

Open nimeng1299 opened 2 months ago

nimeng1299 commented 2 months ago

Can I train a separate eval network myself? If possible, how should I train?

jhonnold commented 2 months ago

You're welcome to do as you please as long as you follow GPL licensing if publishing a release.

I use the Grapheus Trainer which has my model built in. My data is kept locally and I simply don't have the means to upload it, you'd have to find/generate your own dataset.

nimeng1299 commented 2 months ago

You're welcome to do as you please as long as you follow GPL licensing if publishing a release.

I use the Grapheus Trainer which has my model built in. My data is kept locally and I simply don't have the means to upload it, you'd have to find/generate your own dataset.

Thank you for your reply. How do I generate a dataset? Is there a training dataset generator like Stockfish?

nimeng1299 commented 1 month ago

i use nnue-fen-gen-alt to run some .fens, but it doesn't seem to be suitable for Grapheus Trainer. How should I convert this fens?

jhonnold commented 1 month ago

https://github.com/Luecx/fin-tool