evan/nix
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

chore: add desktop

Browse Source
This commit is contained in:
Evan Reichard
2025-12-01 21:09:44 -05:00
parent 23a871a1c9
commit cf5fadf040
3 changed files with 193 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

48
modules/nixos/services/llama-cpp/default.nix
View File

@@ -1,4 +1,9 @@
{ config, pkgs, lib, namespace, ... }:
{ config
, pkgs
, lib
, namespace
, ...
}:
let
inherit (lib) types mkIf mkEnableOption;
inherit (lib.${namespace}) mkOpt;
@@ -40,7 +45,10 @@ in
description = "Download Model";
wantedBy = [ "multi-user.target" ];
before = [ "llama-cpp.service" ];
path = [ pkgs.curl pkgs.coreutils ];
path = [
pkgs.curl
pkgs.coreutils
];
serviceConfig = {
Type = "oneshot";
RemainAfterExit = true;
@@ -85,22 +93,28 @@ in
openFirewall = true;
model = "${modelPath}";
package = (pkgs.llama-cpp.override {
cudaSupport = true;
}).overrideAttrs (oldAttrs: {
cmakeFlags = oldAttrs.cmakeFlags ++ [
"-DGGML_CUDA_ENABLE_UNIFIED_MEMORY=1"
"-DCMAKE_CUDA_ARCHITECTURES=61" # GTX-1070
package =
(pkgs.llama-cpp.override {
cudaSupport = true;
blasSupport = true;
rocmSupport = false;
metalSupport = false;
}).overrideAttrs
(oldAttrs: {
cmakeFlags = oldAttrs.cmakeFlags ++ [
"-DGGML_CUDA_ENABLE_UNIFIED_MEMORY=1"
"-DCMAKE_CUDA_ARCHITECTURES=61" # GTX-1070 / GTX-1080ti
"-DGGML_NATIVE=ON"
# Disable CPU Instructions - Intel(R) Core(TM) i5-3570K CPU @ 3.40GHz
"-DLLAMA_FMA=OFF"
"-DLLAMA_AVX2=OFF"
"-DLLAMA_AVX512=OFF"
"-DGGML_FMA=OFF"
"-DGGML_AVX2=OFF"
"-DGGML_AVX512=OFF"
];
});
# Disable CPU Instructions - Intel(R) Core(TM) i5-3570K CPU @ 3.40GHz
# "-DLLAMA_FMA=OFF"
# "-DLLAMA_AVX2=OFF"
# "-DLLAMA_AVX512=OFF"
# "-DGGML_FMA=OFF"
# "-DGGML_AVX2=OFF"
# "-DGGML_AVX512=OFF"
];
});
extraFlags = [ availableModels.${cfg.modelName}.flag ];
};

7
modules/nixos/virtualisation/podman/default.nix
View File

@@ -1,4 +1,9 @@
{ config, lib, pkgs, namespace, ... }:
{ config
, lib
, pkgs
, namespace
, ...
}:
let
inherit (lib) mkIf;

156
systems/x86_64-linux/lin-va-desktop/default.nix Executable file
View File

@@ -0,0 +1,156 @@
{ namespace
, pkgs
, lib
, ...
}:
let
inherit (lib.${namespace}) enabled;
in
{
system.stateVersion = "25.11";
time.timeZone = "America/New_York";
hardware.nvidia-container-toolkit.enable = true;
nixpkgs.config = {
allowUnfree = true;
packageOverrides = pkgs: {
llama-cpp =
(pkgs.llama-cpp.override {
cudaSupport = true;
blasSupport = true;
rocmSupport = false;
metalSupport = false;
}).overrideAttrs
(oldAttrs: rec {
version = "7213";
src = pkgs.fetchFromGitHub {
owner = "ggml-org";
repo = "llama.cpp";
tag = "b${version}";
hash = "sha256-DVP1eGuSbKd0KxmrDCVAdOe16r1wJ9+RonbrUTgVS7U=";
leaveDotGit = true;
postFetch = ''
git -C "$out" rev-parse --short HEAD > $out/COMMIT
find "$out" -name .git -print0 | xargs -0 rm -rf
'';
};
# Auto CPU Optimizations
cmakeFlags = (oldAttrs.cmakeFlags or [ ]) ++ [
"-DGGML_NATIVE=ON"
"-DGGML_CUDA_ENABLE_UNIFIED_MEMORY=1"
"-DCMAKE_CUDA_ARCHITECTURES=61" # GTX 1070 / GTX 1080ti
];
# Disable Nix's march=native Stripping
preConfigure = ''
export NIX_ENFORCE_NO_NATIVE=0
${oldAttrs.preConfigure or ""}
'';
});
};
};
fileSystems."/mnt/ssd" = {
device = "/dev/disk/by-id/ata-Samsung_SSD_870_EVO_1TB_S6PTNZ0R620739L-part1";
fsType = "exfat";
options = [
"uid=1000"
"gid=100"
"umask=0022"
];
};
networking.firewall = {
allowedTCPPorts = [ 8081 ];
};
# System Config
reichard = {
nix = enabled;
system = {
boot = {
enable = true;
silentBoot = true;
enableSystemd = true;
enableGrub = false;
};
disk = {
enable = true;
diskPath = "/dev/sdc";
};
networking = {
enable = true;
useStatic = {
interface = "enp3s0";
address = "10.0.20.100";
defaultGateway = "10.0.20.254";
nameservers = [ "10.0.20.20" ];
};
};
};
hardware = {
opengl = {
enable = true;
enableNvidia = true;
};
};
services = {
openssh = enabled;
};
virtualisation = {
podman = enabled;
};
};
services.llama-swap = {
enable = true;
openFirewall = true;
settings = {
models = {
# https://huggingface.co/ggml-org/gpt-oss-20b-GGUF/tree/main
"gpt-oss-20b" = {
name = "GPT OSS (20B) - Thinking";
cmd = "${pkgs.llama-cpp}/bin/llama-server --port \${PORT} -m /mnt/ssd/Models/gpt-oss-20b-MXFP4.gguf --ctx-size 128000";
};
# https://huggingface.co/ggml-org/Qwen2.5-Coder-7B-Q8_0-GGUF/tree/main
"qwen2.5-coder-7b-fim" = {
name = "Qwen2.5 Coder (7B) - FIM";
cmd = "${pkgs.llama-cpp}/bin/llama-server -m /mnt/ssd/Models/qwen2.5-coder-7b-q8_0.gguf --fim-qwen-7b-default --port \${PORT}";
};
# https://huggingface.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF/tree/main
"qwen3-30b-a3b-q4-k-m" = {
name = "Qwen3 A3B 2507 (30B) - Instruct";
cmd = "${pkgs.llama-cpp}/bin/llama-server --port \${PORT} -m /mnt/ssd/Models/Qwen3-30B-A3B-Instruct-2507-Q4_K_M.gguf --ctx-size 8192 --temp 0.7 --min-p 0.0 --top-p 0.8 --top-k 20";
};
# https://huggingface.co/unsloth/Qwen3-30B-A3B-Thinking-2507-GGUF/tree/main
"qwen3-30b-a3b-q4-thinking" = {
name = "Qwen3 A3B 2507 (30B) - Thinking";
cmd = "${pkgs.llama-cpp}/bin/llama-server --port \${PORT} -m /mnt/ssd/Models/Qwen3-30B-A3B-Thinking-2507-Q4_K_M.gguf --ctx-size 8192 --temp 0.7 --min-p 0.0 --top-p 0.8 --top-k 20";
};
# https://huggingface.co/unsloth/Qwen3-VL-8B-Instruct-GGUF/tree/main
"qwen3-8b-vision" = {
name = "Qwen3 Vision (8B) - Thinking";
cmd = "${pkgs.llama-cpp}/bin/llama-server --port \${PORT} -m /mnt/ssd/Models/Qwen3-VL-8B-Instruct-UD-Q4_K_XL.gguf --mmproj /mnt/ssd/Models/Qwen3-VL-8B-Instruct-UD-Q4_K_XL_mmproj-F16.gguf --ctx-size 65536 --temp 0.7 --min-p 0.0 --top-p 0.8 --top-k 20";
};
};
};
};
# System Packages
environment.systemPackages = with pkgs; [
btop
git
tmux
vim
llama-cpp
];
}