From 7c217d6ead0c21a3b8e48efece648328afca318a Mon Sep 17 00:00:00 2001
From: Simon Gardling <titaniumtown@proton.me>
Date: Wed, 28 May 2025 21:20:42 -0700
Subject: [PATCH] llama-cpp: use q8 quantization instead of q4

---
 services/llama-cpp.nix | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/services/llama-cpp.nix b/services/llama-cpp.nix
index cb8fdbc..86b4d85 100644
--- a/services/llama-cpp.nix
+++ b/services/llama-cpp.nix
@@ -12,8 +12,8 @@
     enable = true;
     model = builtins.toString (
       pkgs.fetchurl {
-        url = "https://huggingface.co/bartowski/nvidia_AceReason-Nemotron-7B-GGUF/resolve/main/nvidia_AceReason-Nemotron-7B-Q4_0.gguf";
-        sha256 = "27f93349ea88f3c84e53469288ac2ac3f5c985de9f8e00e275870e7e524bb3d8";
+        url = "https://huggingface.co/bartowski/nvidia_AceReason-Nemotron-7B-GGUF/resolve/main/nvidia_AceReason-Nemotron-7B-Q8_0.gguf";
+        sha256 = "0d5eb8b46490af7c097357cb20ad215ebfd30efacedac58bf68a8c7d84e996fc";
       }
     );
     port = service_configs.ports.llama_cpp;