90 lines
2.5 KiB
Nix
90 lines
2.5 KiB
Nix
{ config, lib, inputs, ... }: {
|
|
|
|
options = {
|
|
sysconfig.opts.virtualization.ollama.enable = lib.options.mkOption {
|
|
type = lib.types.bool;
|
|
default = false;
|
|
};
|
|
};
|
|
|
|
config = lib.mkIf config.sysconfig.opts.virtualization.ollama.enable {
|
|
|
|
containers.ollama = {
|
|
|
|
autoStart = true;
|
|
privateNetwork = true;
|
|
hostAddress = "192.168.100.10";
|
|
localAddress = "192.168.100.24";
|
|
|
|
bindMounts = {
|
|
"/dev/nvidia0" = {
|
|
hostPath = "/dev/nvidia0";
|
|
isReadOnly = false;
|
|
};
|
|
"/dev/nvidiactl" = {
|
|
hostPath = "/dev/nvidiactl";
|
|
isReadOnly = false;
|
|
};
|
|
"/dev/nvidia-uvm" = {
|
|
hostPath = "/dev/nvidia-uvm";
|
|
isReadOnly = false;
|
|
};
|
|
"/dev/nvidia-modeset" = {
|
|
hostPath = "/dev/nvidia-modeset";
|
|
isReadOnly = false;
|
|
};
|
|
"/dev/nvidia-uvm-tools" = {
|
|
hostPath = "/dev/nvidia-uvm-tools";
|
|
isReadOnly = false;
|
|
};
|
|
};
|
|
|
|
allowedDevices = [
|
|
{
|
|
node = "/dev/nvidia0";
|
|
modifier = "rw";
|
|
}
|
|
{
|
|
node = "/dev/nvidiactl";
|
|
modifier = "rw";
|
|
}
|
|
{
|
|
node = "/dev/nvidia-uvm";
|
|
modifier = "rw";
|
|
}
|
|
{
|
|
node = "/dev/nvidia-modeset";
|
|
modifier = "rw";
|
|
}
|
|
{
|
|
node = "/dev/nvidia-uvm-tools";
|
|
modifier = "rw";
|
|
}
|
|
];
|
|
|
|
config = {
|
|
|
|
services.ollama = {
|
|
enable = true;
|
|
acceleration = "cuda";
|
|
package = let
|
|
pkgs-us = import inputs.nixpkgs-us {
|
|
system = "x86_64-linux";
|
|
config.allowUnfree = true;
|
|
};
|
|
in pkgs-us.ollama;
|
|
|
|
host = "0.0.0.0";
|
|
|
|
loadModels = [ "llama3.1:8b" ];
|
|
|
|
openFirewall = true;
|
|
};
|
|
|
|
system.stateVersion = "25.05";
|
|
};
|
|
};
|
|
|
|
};
|
|
}
|