diff --git a/llm/llama.cpp b/llm/llama.cpp
index 7593639c..46e12c46 160000
--- a/llm/llama.cpp
+++ b/llm/llama.cpp
@@ -1 +1 @@
-Subproject commit 7593639ce335e8d7f89aa9a54d616951f273af60
+Subproject commit 46e12c4692a37bdd31a0432fc5153d7d22bc7f72
diff --git a/llm/patches/02-clip-log.diff b/llm/patches/02-clip-log.diff
new file mode 100644
index 00000000..34a018e8
--- /dev/null
+++ b/llm/patches/02-clip-log.diff
@@ -0,0 +1,12 @@
+diff --git a/examples/llava/clip.cpp b/examples/llava/clip.cpp
+index e431c7f7..f077e688 100644
+--- a/examples/llava/clip.cpp
++++ b/examples/llava/clip.cpp
+@@ -3,6 +3,7 @@
+ // I'll gradually clean and extend it
+ // Note: Even when using identical normalized image inputs (see normalize_image_u8_to_f32()) we have a significant difference in resulting embeddings compared to pytorch
+ #include "clip.h"
++#include "common.h"
+ #include "log.h"
+ #include "ggml.h"
+ #include "ggml-alloc.h"