From 38d2d4f4ae38bd650e941865e55fb1cff9f43993 Mon Sep 17 00:00:00 2001
From: Harald Hoyer <harald@hoyer.xyz>
Date: Fri, 15 May 2026 07:47:43 +0200
Subject: [PATCH] fix(halo): q6_k with mtp 2

---
 overlays/unstable/default.nix                      | 8 ++++----
 systems/x86_64-linux/halo/llama-server-27B-MTP.nix | 4 ++--
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/overlays/unstable/default.nix b/overlays/unstable/default.nix
index c126ce2..e2be994 100644
--- a/overlays/unstable/default.nix
+++ b/overlays/unstable/default.nix
@@ -22,13 +22,13 @@ final: prev: {
         src = prev.fetchFromGitHub {
           owner = "am17an";
           repo = "llama.cpp";
-          rev = "ebe4fca4b59ef8871bb07c34d148bc37fe57fadd";
-          hash = "sha256-0ogXU/70e7TXtnhN8po0hT+GMEz2wWgob3ah4nBbmuw=";
+          rev = "08b147428e7db0760acda2b4e0bd49f5b2ffe945";
+          hash = "sha256-g74JX/ucsnNI1dWSrrzTUVVbZSg+ShIJfKLbjcD0+ac=";
           postFetch = ''
-            echo -n "ebe4fca4b59ef8871bb07c34d148bc37fe57fadd" > $out/COMMIT
+            echo -n "08b147428e7db0760acda2b4e0bd49f5b2ffe945" > $out/COMMIT
           '';
         };
-        npmDepsHash = "sha256-cV3noOyKmst9vfxyvkCNhihPgwfVGhmPPT4UMloeWZM=";
+        npmDepsHash = "sha256-WaEePrEZ7O/7deP2KJhe0AwiSKYA8HOqETmMHUkmBe0=";
       });
 
   /*
diff --git a/systems/x86_64-linux/halo/llama-server-27B-MTP.nix b/systems/x86_64-linux/halo/llama-server-27B-MTP.nix
index a4ad562..cd58085 100644
--- a/systems/x86_64-linux/halo/llama-server-27B-MTP.nix
+++ b/systems/x86_64-linux/halo/llama-server-27B-MTP.nix
@@ -35,7 +35,7 @@
         "--port 8000"
         "--no-mmap"
         "--n-gpu-layers 99"
-        "-hf unsloth/Qwen3.6-27B-MTP-GGUF:UD-Q8_K_XL"
+        "-hf unsloth/Qwen3.6-27B-MTP-GGUF:Q6_K"
         "--alias halo-8000"
         "--threads 8"
         "--ubatch-size 256"
@@ -46,7 +46,7 @@
         "-c 262144"
         "--fit on"
         "--slot-save-path %C/llama-server/kv-slots-27B-MTP"
-        "--spec-type mtp --spec-draft-n-max 3"
+        "--spec-type draft-mtp --spec-draft-n-max 2"
       ];
       Restart = "on-failure";
       RestartSec = 10;