From c55b4d49f4efd86163e1aa726214be68352def6e Mon Sep 17 00:00:00 2001 From: Liam Barry Date: Wed, 21 May 2025 16:37:35 +0100 Subject: [PATCH] MLECO-6234: Update LLM module to latest revision of llama.cpp dependency Updated function names and commit SHA Signed-off-by: Liam Barry Change-Id: Ib66d8ae1b437e1119e9bc0415a03177f1dbae4ca --- src/cpp/llama_cpp/CMakeLists.txt | 2 +- src/cpp/llama_cpp/include/LlamaImpl.hpp | 1 - src/cpp/llama_cpp/jni/Llama.cpp | 5 ++--- 3 files changed, 3 insertions(+), 5 deletions(-) diff --git a/src/cpp/llama_cpp/CMakeLists.txt b/src/cpp/llama_cpp/CMakeLists.txt index 82a3c31..918be76 100644 --- a/src/cpp/llama_cpp/CMakeLists.txt +++ b/src/cpp/llama_cpp/CMakeLists.txt @@ -25,7 +25,7 @@ set(LLAMA_GIT_URL "https://github.com/ggerganov/llama.cpp.git" CACHE STRING "Git URL for llama.cpp repo") -set(LLAMA_GIT_SHA "c5d91a7" +set(LLAMA_GIT_SHA "a4090d1" CACHE STRING "Git commit SHA for llama.cpp repo") diff --git a/src/cpp/llama_cpp/include/LlamaImpl.hpp b/src/cpp/llama_cpp/include/LlamaImpl.hpp index ff362e5..0982c1d 100644 --- a/src/cpp/llama_cpp/include/LlamaImpl.hpp +++ b/src/cpp/llama_cpp/include/LlamaImpl.hpp @@ -11,7 +11,6 @@ #include "llama.h" #include "common.h" #include "LLM.hpp" -#include "llama-sampling.h" #define LOG_INF(...) do { fprintf(stdout, __VA_ARGS__); } while (0) diff --git a/src/cpp/llama_cpp/jni/Llama.cpp b/src/cpp/llama_cpp/jni/Llama.cpp index a0bf6b3..6f8c980 100644 --- a/src/cpp/llama_cpp/jni/Llama.cpp +++ b/src/cpp/llama_cpp/jni/Llama.cpp @@ -45,7 +45,7 @@ JNIEXPORT jlong JNICALL Java_com_arm_llm_Llama_loadModel(JNIEnv *env, jobject, j JNIEXPORT void JNICALL Java_com_arm_llm_Llama_kvCacheClear(JNIEnv, jobject, jlong contextPtr) { - llama_kv_cache_clear(reinterpret_cast(contextPtr)); + llama_kv_self_clear(reinterpret_cast(contextPtr)); } /** @@ -57,8 +57,7 @@ Java_com_arm_llm_Llama_kvCacheClear(JNIEnv, jobject, jlong contextPtr) JNIEXPORT void JNICALL Java_com_arm_llm_Llama_kvCacheSeqRm(JNIEnv, jobject, jlong contextPtr, jint start_pos, jint last_pos) { - - llama_kv_cache_seq_rm(reinterpret_cast(contextPtr), -1, start_pos, last_pos); + llama_kv_self_seq_rm(reinterpret_cast(contextPtr), -1, start_pos, last_pos); } /** -- GitLab