| 14 May 2025 |
connor (burnt/out) (UTC-8) | I also still owe you a review of https://github.com/NixOS/nixpkgs/pull/406740! | 08:15:09 |
SomeoneSerge (back on matrix) | Yea I shut down myself for like 12 hours. Coffee time! | 11:48:56 |
| 16 May 2025 |
SomeoneSerge (back on matrix) | connor (he/him) (UTC-7): I guess it's trivial but I only now confirmed that you actually can _file = jsonPath in evalModules! | 15:25:22 |
connor (burnt/out) (UTC-8) | Remind me, _file (just?) helps with debugging because it tells you where configurations came from on disk, right? I remember seeing it set by lib.importApply (introduced by Robert from flake-parts IIRC) | 15:30:18 |
SomeoneSerge (back on matrix) | The option licenses.distribution_path."CUDA Toolkit" is defined both null and not null, in ... | 15:41:27 |
SomeoneSerge (back on matrix) | ... could be somethin-something-undefined (because of a lambda) or could be a path | 15:42:16 |
SomeoneSerge (back on matrix) | * ... could be somethin-something-undefined (because of a lambda or let-in) or could be a path | 15:42:27 |
hexa (UTC+1) | working on migrating python3 to 3.13 | 22:28:11 |
| 17 May 2025 |
| terrorjack set a profile picture. | 08:53:50 |
| 19 May 2025 |
SomeoneSerge (back on matrix) | connor (he/him) (UTC-7): I started (1) switching to a column-oriented format in the last push, and I also (2) attrsOf (enum [ 1 ]) instead of listOf to enforce uniqueness... this is over-engineered isn't it? | 02:05:47 |
| oak 🏳️🌈♥️ changed their display name from oak 🫱⭕🫲 to oak. | 10:59:27 |
SomeoneSerge (back on matrix) | * connor (he/him) (UTC-7): I started (1) switching to a column-oriented format in the last push, and I also (2) use attrsOf (enum [ 1 ]) instead of listOf to enforce uniqueness... this is over-engineered isn't it? | 02:06:08 |
| oak 🏳️🌈♥️ changed their display name from oak to oak 🏳️🌈♥️. | 11:01:11 |
hexa (UTC+1) | tensorflow is still disabled on 3.13 | 15:41:14 |
hexa (UTC+1) | * tensorflow-bin is still disabled on 3.13 | 15:41:17 |
hexa (UTC+1) | yeah, so 2.20 nightlies support 3.13 since april | 15:53:16 |
hexa (UTC+1) | they really dropped the ball here | 15:53:20 |
| 20 May 2025 |
connor (burnt/out) (UTC-8) | too much brain ouch | 04:11:11 |
connor (burnt/out) (UTC-8) | talk about it tomorrow if you're still good for our chat? | 04:12:13 |
breakds | Can someone here help take a look at this PR: https://github.com/NixOS/nixpkgs/pull/408555 ? It is the packaging of flashinfer, a library for transformer inference. The produced package works locally, but I am not very confident that I did everything correctly in the packaging - it might be good to get looked by an expert. Thanks a lot! | 15:59:56 |
connor (burnt/out) (UTC-8) | SomeoneSerge (UTC+U[-12,12]): check out the latest commit on https://github.com/NixOS/nixpkgs/pull/406531, should have addressed everything we talked about this morning I still need to review yours, should be able to later today 🫠 | 17:48:26 |
SomeoneSerge (back on matrix) | I fell asleep after the 3rd coffee | 19:58:36 |
| 21 May 2025 |
Kevin Mittman (UTC-8) | arches 👀 | 01:40:13 |
connor (burnt/out) (UTC-8) | I should be able to do the second half of the review on your PR tomorrow morning. Gonna try to be at work before 6:30 again. | 04:44:41 |
SomeoneSerge (back on matrix) | What's your opinion on this? | 13:36:17 |
| 23 May 2025 |
connor (burnt/out) (UTC-8) | Okay hoping I get a sudden burst of productivity and get through my backlog | 12:29:14 |
SomeoneSerge (back on matrix) | connor (he/him) (UTC-7): I'm finally beginning to delete things! | 19:52:09 |
| 24 May 2025 |
ereslibre | Hi! If you have some time I have a couple PRs waiting review/merge:
- Allow to provide CSV files for Nvidia-ctk (fixes jetson devices): https://github.com/NixOS/nixpkgs/pull/401840
- Fix nvidia-ctk for the Nvidia docker runtime (deprecated): https://github.com/NixOS/nixpkgs/pull/407290
Thanks :) | 07:46:45 |
little_dude | Download x | 09:07:51 |
little_dude | Hello, this was a long time ago, but I'm finally back to trying to run ollama :D
saxpy doesn't work. I used this flake:
{
description = "CUDA saxpy test";
inputs.nixpkgs.url = "nixpkgs";
outputs =
{ self, nixpkgs }:
{
devShell.x86_64-linux =
let
pkgs = import nixpkgs {
system = "x86_64-linux";
config.allowUnfree = true; # Required for CUDA
};
in
pkgs.mkShell {
name = "cuda-saxpy-shell";
buildInputs = [
pkgs.cudaPackages.saxpy
pkgs.cudaPackages.cudatoolkit
];
shellHook = ''
export CUDA_PATH=${pkgs.cudatoolkit}
export EXTRA_LDFLAGS="-L/lib -L${pkgs.linuxPackages.nvidia_x11}/lib"
export EXTRA_CCFLAGS="-I/usr/include"
# Should I set this?
# export LD_LIBRARY_PATH=${pkgs.cudaPackages.cudatoolkit.lib}/lib:$LD_LIBRARY_PATH
'';
};
};
}
I'm running in the same(?) initialization error I think (see the log file attached) for LD_DEBUG=libs saxpy.
The output of nvidia-smi:
{
description = "CUDA saxpy test";
inputs.nixpkgs.url = "nixpkgs";
outputs =
{ self, nixpkgs }:
{
devShell.x86_64-linux =
let
pkgs = import nixpkgs {
system = "x86_64-linux";
config.allowUnfree = true; # Required for CUDA
};
in
pkgs.mkShell {
name = "cuda-saxpy-shell";
buildInputs = [
pkgs.cudaPackages.saxpy
pkgs.cudaPackages.cudatoolkit
];
shellHook = ''
export CUDA_PATH=${pkgs.cudatoolkit}
export EXTRA_LDFLAGS="-L/lib -L${pkgs.linuxPackages.nvidia_x11}/lib"
export EXTRA_CCFLAGS="-I/usr/include"
# Should I set this?
# export LD_LIBRARY_PATH=${pkgs.cudaPackages.cudatoolkit.lib}/lib:$LD_LIBRARY_PATH
'';
};
};
}
| 09:08:34 |