push
This commit is contained in:
@@ -1,10 +1,27 @@
|
||||
{ ... }: {
|
||||
den.aspects.ollama = {
|
||||
nixos = {
|
||||
services.ollama = {
|
||||
den.aspects.llama = {
|
||||
nixos = { pkgs, ... }: {
|
||||
services.llama-cpp = {
|
||||
enable = true;
|
||||
package = pkgs.llama-cpp.override { cudaSupport = true; };
|
||||
settings = {
|
||||
host = "0.0.0.0";
|
||||
port = 8080;
|
||||
ctx-size = 65536;
|
||||
|
||||
loadModels = [ "llama3.2:3b" "deepseek-r1:1.5b" "deepseek-r1:8b" "qwen2.5-coder:7b" "qwen3.5:9b" "aratan/qwen3.5-agent-multimodal:9b" "qwen3-coder:30b" ];
|
||||
models-preset = (pkgs.formats.ini { }).generate "llama-cpp-models-preset.ini" {
|
||||
"Qwen3.6-35B-A3B" = {
|
||||
hf-repo = "unsloth/Qwen3.6-35B-A3B-GGUF";
|
||||
hf-file = "Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf";
|
||||
alias = "qwen3.6-35b-a3b";
|
||||
temp = "1.0";
|
||||
top-p = "0.95";
|
||||
top-k = "20";
|
||||
n-gpu-layers = 80;
|
||||
ctx-size = 65536;
|
||||
};
|
||||
};
|
||||
};
|
||||
};
|
||||
};
|
||||
};
|
||||
|
||||
Reference in New Issue
Block a user