Guidelines for pre-trained ML model weight binaries (Was re: Where should we put machine learning model parameters?)

I am uncomfortable with including ML models without their training data available. It is possible to hide backdoors in them. https://www.quantamagazine.org/cryptographers-show-how-to-hide-invisible-backdoors-in-ai-20230302/

GOOPS-less Shepherd

Hello! I’d like to release the Shepherd 0.10.0 in a few weeks at most, with the hope that it’ll be the last stable series before 1.0, which would be released in a few months. As part of this, I’d like to clean up the API, which includes removing the dependency on GOOPS. The Shepherd had been usi

Commits and bug closing (was: something else)

Hello, Am Wed, Apr 05, 2023 at 09:19:43AM -0700 schrieb Felix Lechner: > Do we have a hook that closes such bugs automatically via instructions > in commit messages? > If not, I'd be happy to look into writing such a thing. It would also > help to tie commits to bug reports, which can be good for

how to deal with large dataset? (was Re: Where should we put machine learning model parameters ?)

Hi, Well, we already discussed in GWL context where to put “large” data set without reaching a conclusion. Having “large” data set inside the store is probably not a good idea. But maybe these data of models are not that “large” to worry about the store. On lun., 03 avril 2023 at 18:48, Nicola

Re: OpenJDK in core-updates

Am Wed, Apr 05, 2023 at 02:26:07PM +0200 schrieb Andreas Enge: > Next problem, hopefully fixed by the following patch: Funny observation, the problems appear in even releases... So OpenJDK 12, 14 and 16 contain bugs that were fixed in 11, 13 and 15! As far as I can tell, this is fixed now on core-

Re: Guidelines for pre-trained ML model weight binaries (Was re: Where should we put machine learning model parameters?)

Hi, On Thu, 6 Apr 2023 at 15:41, Kyle wrote: > I have only seen situations where the optimization is "too entailed with > randomness" when models are trained on proprietary GPUs with specific > settings. Otherwise, pseudo-random seeds are perfectly sufficient to remove > the indeterminism. F

Re: Guidelines for pre-trained ML model weight binaries (Was re: Where should we put machine learning model parameters?)

>Since it is computing, we could ask about the bootstrap of such >generated data. I think it is a slippery slope because it is totally >not affordable to re-train for many cases: (1) we would not have the >hardware resources from a practical point of view,, (2) it is almost >impossible to tac

Re: intrinsic vs extrinsic identifier: toward more robustness?

Hi, On jeu., 16 mars 2023 at 18:45, Ludovic Courtès wrote: >> For sure, we have to fix the holes and bugs. :-) However, I am asking >> what we could add for having more robustness on the long term. > Sources (fixed-output derivations) are already content-addressed, by > definition (I prefer “c

Re: [internship]GSoC proposal review period begins today

Hi Gábor, On Tue, 04 Apr 2023 at 22:14, Gábor Boskovits wrote: > Don't worry, we have some. I am going to have a look now, as it should now > be final. Cool! Thank you. Cheers, simon

Re: Guidelines for pre-trained ML model weight binaries (Was re: Where should we put machine learning model parameters?)

Hi, On Mon, 03 Apr 2023 at 18:07, Ryan Prior wrote: > Hi there FSF Licensing! (CC: Guix devel, Nicholas Graves) This morning > I read through the FSDG to see if it gives any guidance on when > machine learning model weights are appropriate for inclusion in a free > system. It does not seem to of