!eWOErHSaiddIbsUNsJ:nixos.org

NixOS CUDA

303 Members
CUDA packages maintenance and support in nixpkgs | https://github.com/orgs/NixOS/projects/27/ | https://nixos.org/manual/nixpkgs/unstable/#cuda59 Servers

You have reached the beginning of time (for this room).


SenderMessageTime
6 Oct 2025
@ss:someonex.netSomeoneSerge (matrix works sometimes)Well the real issue here is that lol why physical realizations of aterm drvs in the first place19:44:03
7 Oct 2025
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)Using toJSON calls addToStore on paths before serializing so they’re valid; I guess I could try using unsafeDiscardWhatever before serializing to JSON to see if that’s enough to prevent realization01:32:05
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8) There was a --read-only flag I had forgotten about 03:47:47
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)

Okay, it's way faster now and doesn't need a ramdisk.
Where as previously it would take about 1m40s on my i9-13900k now it's taking about 20s to do

time nix build -L ".#reports.x86_64-linux.pkgs-cuda-post^*" --builders '' --rebuild
03:53:40
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)For generating a single report htop shows it was taking about 18% of my RAM (less than 20GB)03:57:11
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)To clarify, it wasn’t enough to use unsafeDiscardStringContext because the derivation was instantiated as soon as drvPath was evaluated, even before toJSON. Using the read-only argument (which is different than the store URI query parameter of the same name lmao) avoids the instantiating.06:54:57
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)https://cuda-index.someonex.net15:40:22
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)https://forge.someonex.net/else/sidx15:40:26
@ss:someonex.netSomeoneSerge (matrix works sometimes) connor (he/him) (UTC-7): ping about azure stuff again 19:30:18
@ss:someonex.netSomeoneSerge (matrix works sometimes)Hey, we've been getting more experience with infra and hydra now, and I think like ephemeral builders are becoming more and more relevant. May I ask you to elaborate?19:34:01
@ss:someonex.netSomeoneSerge (matrix works sometimes) * connor (he/him) (UTC-7): ping about azure images again 19:34:34
@ss:someonex.netSomeoneSerge (matrix works sometimes) MI25s to test ROCm? CC Lun
https://docs.azure.cn/en-us/virtual-machines/sizes/gpu-accelerated/nv-family
19:39:24
@lt1379:matrix.orgLunToo old19:39:46
@ss:someonex.netSomeoneSerge (matrix works sometimes)* Hey, we've been getting more experience with infra and hydra now, and I feel like ephemeral builders are becoming more and more relevant. May I ask you to elaborate?19:39:48
@ss:someonex.netSomeoneSerge (matrix works sometimes)mx250 only?19:40:33
@lt1379:matrix.orgLungfx906 is barely cared for by upstream so not sure they'd be great for automated testing If we have credits that would otherwise go to waste and that's the only option then maybe worth it?19:41:00
@lt1379:matrix.orgLungfx90a (MI210/250) is the oldest instinct option upstream seems to actually be paying attention to19:41:26
@ss:someonex.netSomeoneSerge (matrix works sometimes)Relatable: azure offers the right hw for us, but I'm not confident we can utilize it efficiently enough yet19:44:47

Show newer messages


Back to Room ListRoom Version: 9