From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:470:142:3::10]:57910) by lists.gnu.org with esmtp (Exim 4.86_2) (envelope-from ) id 1hanwO-0004EN-G3 for guix-patches@gnu.org; Tue, 11 Jun 2019 17:04:30 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1hanpK-0002ku-58 for guix-patches@gnu.org; Tue, 11 Jun 2019 16:57:08 -0400 Received: from debbugs.gnu.org ([209.51.188.43]:46713) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1hanpG-0002iQ-J6 for guix-patches@gnu.org; Tue, 11 Jun 2019 16:57:04 -0400 Received: from Debian-debbugs by debbugs.gnu.org with local (Exim 4.84_2) (envelope-from ) id 1hanpG-0007BA-DD for guix-patches@gnu.org; Tue, 11 Jun 2019 16:57:02 -0400 Subject: [bug#36071] [PATCH] gnu: Add leela-zero. Resent-Message-ID: References: <20190603130611.32142-1-glv@posteo.net> <87v9xkyrzz.fsf@gnu.org> <20190605192625.3c567ce7@yamatai> <87tvd3wyhx.fsf@gnu.org> From: Ricardo Wurmus In-reply-to: <87tvd3wyhx.fsf@gnu.org> Date: Tue, 11 Jun 2019 22:55:45 +0200 Message-ID: <87lfy7alke.fsf@elephly.net> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: guix-patches-bounces+kyle=kyleam.com@gnu.org Sender: "Guix-patches" To: Ludovic =?UTF-8?Q?Court=C3=A8s?= Cc: Guillaume LE VAILLANT , 36071@debbugs.gnu.org Ludovic Court=C3=A8s writes: >> Ludovic Court=C3=A8s a =C3=A9crit : >> >>> Does the game run at all if we don=E2=80=99t download additional network >>> weight data? >>> >>> There=E2=80=99s a debate as to whether trained neural network parameter= s can >>> be considered =E2=80=9Csource=E2=80=9D: . >> >> The game needs neural network weights to run. The user must download >> a file with the weights (usually from >> , which is updated daily I think) >> and put it in '$HOME/.local/share/leela-zero/'. >> >> Do you think some weigths (e.g. taken from >> , for reproducibility) should be >> embedded in the package? > > My point is about whether these trained neural network data are > something that we could distribute per the FSDG. > > The LWN article above mentions =E2=80=98leela-zero=E2=80=99. In Debian (= which has a > different policy), =E2=80=98leela-zero=E2=80=99 has apparently been accep= ted, but > without the data, and with a link to said data (as in your original > patch): > > https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=3D903634 > > Perhaps we could do the same, but I=E2=80=99d like to hear what others th= ink. We need to ensure that the software necessary to train the networks is included. Is this the case? Distributing the training data is not an option, as I see it, because the files are very large. I don=E2=80=99t know about the networks that are offered for download at https://leela.online-go.com/networks/. These are files of very different sizes, and it is neither clear what license applies nor how they were arrived at. They are zipped up magic numbers. Arguably a similar statement could be made about impulse response files for audio filters (such as convolution reverbs). While we offer convolution filters (such as the =E2=80=9Cir=E2=80=9D package), we don=E2= =80=99t include data files. Had I not made the link to network weights for ML models I would have considered impulse response files no different from e.g. game assets. Now I=E2=80=99m not so sure any more. I don=E2=80=99t really know what it really means for data to be =E2=80=9Cnon-functional=E2=80=9D. Back to this patch: I think it=E2=80=99s fine to accept it as long as the software necessary for training is included. -- Ricardo