Package Details: llama.cpp-cuda b10082-1

Git Clone URL: https://aur.archlinux.org/llama.cpp-cuda.git (read-only, click to copy)
Package Base: llama.cpp-cuda
Description: Port of Facebook's LLaMA model in C/C++ (with NVIDIA CUDA optimizations)
Upstream URL: https://github.com/ggml-org/llama.cpp
Licenses: MIT
Conflicts: ggml, libggml, llama.cpp
Provides: llama.cpp
Submitter: txtsd
Maintainer: fabse
Last Packager: fabse
Votes: 22
Popularity: 3.04
First Submitted: 2024-10-26 20:17 (UTC)
Last Updated: 2026-07-22 07:35 (UTC)

Dependencies (19)

Required by (13)

Sources (3)

Pinned Comments

txtsd commented on 2024-10-26 20:17 (UTC) (edited on 2024-12-06 14:15 (UTC) by txtsd)

Alternate versions

llama.cpp
llama.cpp-vulkan
llama.cpp-sycl-fp16
llama.cpp-sycl-fp32
llama.cpp-cuda
llama.cpp-cuda-f16
llama.cpp-hip

Latest Comments

1 2 3 4 5 6 .. 11 Next › Last »

shaybox commented on 2026-07-01 18:29 (UTC)

Since you already have to manage a module preset file to specify per-model settings without manually editing the llama.cpp args and restarting the service each time, and all args can be set in that file, I just replaced the EnvironmentFile with the model-preset file.

ExecStart=/usr/bin/llama-server --models-preset /etc/conf.d/llama.cpp

$ systemctl edit llama.cpp.service

[Service]
EnvironmentFile=
ExecStart=
ExecStart=/usr/bin/llama-server --models-preset /etc/conf.d/llama.cpp

version = 1

[*]
cache-type-k = q8_0
cache-type-v = q8_0

[unsloth/Qwen3.6-27B-MTP-GGUF:IQ4_XS]
model = /home/shaybox/.lmstudio/models/unsloth/Qwen3.6-27B-MTP-GGUF/Qwen3.6-27B-IQ4_XS.gguf

[unsloth/Qwen3.6-35B-A3B-MTP-GGUF:IQ4_XS]
model = /home/shaybox/.lmstudio/models/unsloth/Qwen3.6-35B-A3B-MTP-GGUF/Qwen3.6-35B-A3B-UD-IQ4_XS.gguf

fabse commented on 2026-06-29 11:33 (UTC)

while I don't have anything against the workaround, I also would prefer to not do any workaround at all lol, just as to maintain less

that said, @seryogakovalyov's pr got merged a couple of hours ago! b9840 (and hopefully omward) should work given my limited testing (b9839 has the fix, but didn't work with this package because of the upward walk git workaround, which I've since removed; CI is currently running for b9840)

dylanhall commented on 2026-06-29 05:49 (UTC)

can we implement the workaround? that repo has over 1.1k pull requests, i don't think the fix is going to be merged very soon, implementing the workaround the PKGBUILD seems logical

seryogakovalyov commented on 2026-06-23 12:51 (UTC) (edited on 2026-06-23 13:15 (UTC) by seryogakovalyov)

I’d keep the AUR root .gitignore as-is and just add a temporary downstream workaround in prepare():

prepare() {
  ln -sf "${_pkgname}-${pkgver}" llama.cpp
  mkdir -p "${_pkgname}/.git"

  printf '%s\n' '!*' > "${_pkgname}/tools/ui/src/.gitignore"

  if ! grep -q 'src/.gitignore' "${_pkgname}/tools/ui/sources.cmake"; then
    sed -i '/package-lock.json/a\    src/.gitignore' "${_pkgname}/tools/ui/sources.cmake"
  fi
}

fabse commented on 2026-06-23 10:28 (UTC)

oof, again... yeah same for me, also with the llama.cpp and llama.cpp-vulkan packages

thanks for opening a PR with a fix too, hopefully that's merged soon! in the meantime, should I just go ahead and update the package to remove .gitignore? I see no risk doing so, but still an unusual change

seryogakovalyov commented on 2026-06-21 21:13 (UTC) (edited on 2026-06-22 04:24 (UTC) by seryogakovalyov)

I found the root cause of the broken Web UI and opened an upstream PR with a fix: https://github.com/ggml-org/llama.cpp/pull/24879

Local workaround for AUR users:

git clone https://aur.archlinux.org/llama.cpp-cuda.git

cd llama.cpp-cuda

rm .gitignore

makepkg -si

After that the Web UI should render normally again.

seryogakovalyov commented on 2026-06-21 08:45 (UTC) (edited on 2026-06-21 10:42 (UTC) by seryogakovalyov)

The UI is broken for me starting from b9716-1. Does anyone know how to fix it?

fabse commented on 2026-06-20 20:41 (UTC)

ahh otherwise the ui will fail to build, setting '''BUILD_UI''' will call npm build: https://github.com/ggml-org/llama.cpp/blob/master/scripts/ui-assets.cmake

manylegged commented on 2026-06-20 19:23 (UTC)

Why is npm in makedepends? I just built it locally without npm installed. Especially in light of recent AUR attacks it seems prudent not to require this extraneously

1 2 3 4 5 6 .. 11 Next › Last »