Re: training dataset for gnubg.weights

[Stefano Zacchiroli]

This is awesome, thank you. I was going to follow up to mention that,
offlist, Øystein Schønning-Johansen has shared with me the training data
and offer to host it somewhere, but it looks like that's not needed,
great!

Would it be possible to add these links somewhere in the GnuBG doc? With
all the ongoing debated about free/open AI, I think "where is the
training data" is going to become a FAQ-level question. If a patch for
this would be welcome, shout and I'll be happy to draft something.

I'm going to have a look at the Git repo to see if I can figure out how
to (re)do the training.

Thanks again,
Cheers

On Thu, Oct 31, 2024 at 03:17:34PM +0100, Philippe Michel wrote:
> On Wed, Oct 30, 2024 at 10:57:44AM -0400, Stefano Zacchiroli wrote:
> 
> > I'm looking for the training dataset and training code/pipeline used to
> > generated the gnubg.weights shipped from the Git repo at [1]. Does
> > anyone know where to find them?
> 
> The training data can be found at 
> https://alpha.gnu.org/gnu/gnubg/nn-training/
> and the training code at 
> https://git.savannah.gnu.org/cgit/gnubg/gnubg-nn.git/
> 
> Note that the latter is only the training tools. Documentation is minimal 
> and there is no simple recipe to reproduce the current results.
> 

-- 
Stefano Zacchiroli . zack@upsilon.cc . https://upsilon.cc/zack  _. ^ ._
Full professor of Computer Science              o     o   o     \/|V|\/
Télécom Paris, Polytechnic Institute of Paris     o     o o    </>   <\>
Co-founder & CSO Software Heritage            o o o     o       /\|^|/\
Mastodon: https://mastodon.xyz/@zacchiro                        '" V "'