!eWOErHSaiddIbsUNsJ:nixos.org

NixOS CUDA

314 Members
CUDA packages maintenance and support in nixpkgs | https://github.com/orgs/NixOS/projects/27/ | https://nixos.org/manual/nixpkgs/unstable/#cuda62 Servers

Load older messages


SenderMessageTime
12 Apr 2026
@kaya:catnip.eekaya 𖤐 changed their profile picture.05:53:35
@leona:leona.isleona changed their profile picture.12:15:52
14 Apr 2026
@danielrf:matrix.orgdanielrf connor (burnt/out) (UTC-8): TensorRT support in onnxruntime. It pretty much just works if you turn on support for it: https://github.com/NixOS/nixpkgs/pull/509976 13:40:27
15 Apr 2026
@jonathandekhtiar:matrix.orgJohn D joined the room.22:22:58
16 Apr 2026
@ccicnce113424:matrix.orgccicnce113424Redacted or Malformed Event01:14:52
@ccicnce113424:matrix.orgccicnce113424 * I'm using the latest version of nixos-unstable and noticed that Firefox isn't being cached. I checked your Hydra and found that the last successful evaluation for cuda-packages-unstable occurred on April 1; the most recent evaluation processes all ended prematurely. 01:29:07
@ccicnce113424:matrix.orgccicnce113424* I checked your Hydra and found that the last successful evaluation for cuda-packages-unstable occurred on April 1; the most recent evaluation processes all ended prematurely.01:29:34
@jaredmontoya:matrix.orgjaredmontoyaI tried to update to the latest nixos-unstable but thunderbird seems to not be in cache.nixos-cuda.org (and I can't build it because it sems to need more than 64GB RAM to build) did anyone else experience this?13:33:23
@hexa:lossy.networkhexa (UTC+1)

more than 64 GB

13:34:23
@hexa:lossy.networkhexa (UTC+1)that seems unlikely13:34:26
@hexa:lossy.networkhexa (UTC+1)https://hydra.nixos-cuda.org/jobset/nixos-cuda/cuda-packages-unstable#tabs-evaluations13:35:18
@hexa:lossy.networkhexa (UTC+1)the last eval was two weeks ago13:35:24
@sporeray:matrix.orgRobbie Buxton
In reply to @hexa:lossy.network
that seems unlikely
Idk I’ve oomed on packages with over 1TB of ram
13:36:51
@hexa:lossy.networkhexa (UTC+1)wat\13:37:01
@hexa:lossy.networkhexa (UTC+1)Redacted or Malformed Event13:37:03
@gsaurel:laas.frnim65sbut you build with how many cores?13:37:23
@hexa:lossy.networkhexa (UTC+1)a) you can run out of ram with exccessive cores13:37:29
@gsaurel:laas.frnim65s* but you build with how many cores ?13:37:31
@hexa:lossy.networkhexa (UTC+1)b) without excessive cores the most ram usage is during linking13:37:38
@hexa:lossy.networkhexa (UTC+1)I'd say firefox takes at most 40 GB to link13:37:55
@sporeray:matrix.orgRobbie Buxton I was using nixbuild.net and it kept sizing it up and giving it more cores, and yeah the solution was reducing cores obviously 13:38:02
@hexa:lossy.networkhexa (UTC+1)that's not friendly at all, but does fit on my desktop still13:38:12
@hexa:lossy.networkhexa (UTC+1)but that means your core/ram balance was way off13:38:28
@sporeray:matrix.orgRobbie Buxton
In reply to @hexa:lossy.network
but that means your core/ram balance was way off
How so?
13:39:20
@hexa:lossy.networkhexa (UTC+1)you have a base need of RAM per core13:39:51
@hexa:lossy.networkhexa (UTC+1)just to run the build on it13:39:58
@hexa:lossy.networkhexa (UTC+1)and if you oom the total cores probably needed more ram than what was provided13:40:17
@hexa:lossy.networkhexa (UTC+1)and the solution would have been to get fewer cores with more ram instead13:40:30
@sporeray:matrix.orgRobbie BuxtonIdk I can’t get more than 1.6 TB of ram on AWS 😁13:40:42
@hexa:lossy.networkhexa (UTC+1)now I don't know anything about how nixbuild.net does scaling13:40:56

Show newer messages


Back to Room ListRoom Version: 9