IRC channel logs
2021-06-08.log
back to list of logs
<civodul>what channels do we have that provide machine learning things? <rekado>the “guix” channel has a (gnu packages machine-learning) module ;) <rekado>I have no secret machine learning channel, unfortunately <civodul>yeah i was wondering if there was a quick'n'dirty tensorflow package somewhere, for instance <rekado>we have tensorflow 1 in Guix; I haven’t recently tried to package tensorflow 2. <PurpleSym>I remember efraim was working on a bazel package? <efraim>I walked away from it, but I can send you what I had <efraim>I don't remember the state they're in, but it wasn't "working" <efraim>CUDA.jl hard depends on CUDA for functionality, but doesn't need it to install the package <efraim>I think I can just hard drop it while upstreaming. It was only for the tests anyway <civodul>just gave a 10mn (!) talk about Guix to AI developers at work <rekado>you’re all so busy presenting! Truly admirable. <rekado>I’m preparing the migration from the slow NFS appliance to a new HA NFS setup (where guix-daemon has local access to the disks) <rekado>the move is going to happen next Wednesday <civodul>are there still open issues with setting up Guix? <civodul>would be nice if you could test before it's in production <rekado>this is all going to be a little difficult because I have no write access to the configuration management system for the cluster <rekado>so all I can do is ask other people to make the changes I’m blindly planning <rekado>but I hope I can test drive this on an isolated cluster node later this week <jra>Anyone know of any docs/guides on how to create relocatable mpi-app packs to run on non-Guix HPC systems? I've gotten as far as building an example mpi app, ring_c, making a relocatable tarball. I haven't been able to get it to run on an HPC system. <civodul>jra: hi! could you share the specifics? <civodul>on one system i had troubles related to the name service switch (NSS)