chore(deps): bump llama.cpp to b34c859146630dff136943abc9852ca173a7c9d6 (#5323)

chore(deps): bump llama.cpp to 'b34c859146630dff136943abc9852ca173a7c9d6'

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

Makefile

@@ -6,7 +6,7 @@ BINARY_NAME=local-ai
 DETECT_LIBS?=true
 
 # llama.cpp versions
-CPPLLAMA_VERSION?=9fdfcdaeddd1ef57c6d041b89cd8fb7048a0f028
+CPPLLAMA_VERSION?=b34c859146630dff136943abc9852ca173a7c9d6
 
 # whisper.cpp version
 WHISPER_REPO?=https://github.com/ggerganov/whisper.cpp

backend/cpp/llama/patches/01-llava.patch

@@ -1,7 +1,7 @@
-diff --git a/tools/llava/clip.cpp b/tools/llava/clip.cpp
+diff --git a/tools/mtmd/clip.cpp b/tools/mtmd/clip.cpp
 index 3cd0d2fa..6c5e811a 100644
---- a/tools/llava/clip.cpp
-+++ b/tools/llava/clip.cpp
+--- a/tools/mtmd/clip.cpp
++++ b/tools/mtmd/clip.cpp
 @@ -2608,7 +2608,7 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima
                  struct ggml_tensor * patches = ggml_graph_get_tensor(gf, "patches");
                  int* patches_data = (int*)malloc(ggml_nbytes(patches));

backend/cpp/llama/prepare.sh

@@ -20,9 +20,9 @@ fi
 
 ## XXX: In some versions of CMake clip wasn't being built before llama.
 ## This is an hack for now, but it should be fixed in the future.
-cp -rfv llama.cpp/tools/llava/clip.h llama.cpp/tools/grpc-server/clip.h
-cp -rfv llama.cpp/tools/llava/clip-impl.h llama.cpp/tools/grpc-server/clip-impl.h
-cp -rfv llama.cpp/tools/llava/llava.cpp llama.cpp/tools/grpc-server/llava.cpp
+cp -rfv llama.cpp/tools/mtmd/clip.h llama.cpp/tools/grpc-server/clip.h
+cp -rfv llama.cpp/tools/mtmd/clip-impl.h llama.cpp/tools/grpc-server/clip-impl.h
+cp -rfv llama.cpp/tools/mtmd/llava.cpp llama.cpp/tools/grpc-server/llava.cpp
 echo '#include "llama.h"' > llama.cpp/tools/grpc-server/llava.h
-cat llama.cpp/tools/llava/llava.h >> llama.cpp/tools/grpc-server/llava.h
-cp -rfv llama.cpp/tools/llava/clip.cpp llama.cpp/tools/grpc-server/clip.cpp
+cat llama.cpp/tools/mtmd/llava.h >> llama.cpp/tools/grpc-server/llava.h
+cp -rfv llama.cpp/tools/mtmd/clip.cpp llama.cpp/tools/grpc-server/clip.cpp

backend/cpp/llama/utils.hpp

@@ -11,7 +11,7 @@
 
 #include "json.hpp"
 
-#include "../llava/clip.h"
+#include "../mtmd/clip.h"
 
 using json = nlohmann::json;