llama-cpp-pro 0.2.2

package/android/src/main/CMakeLists-arm64.txt

@@ -0,0 +1,138 @@
+cmake_minimum_required(VERSION 3.10)
+
+project(llama-cpp)
+
+set(CMAKE_CXX_STANDARD 17)
+set(LLAMACPP_LIB_DIR ${CMAKE_SOURCE_DIR}/../../../cpp)
+
+include_directories(
+    ${LLAMACPP_LIB_DIR}
+    ${LLAMACPP_LIB_DIR}/ggml-cpu
+    ${LLAMACPP_LIB_DIR}/tools/mtmd
+    ${LLAMACPP_LIB_DIR}/vendor/cpp-httplib
+)
+
+set(
+    SOURCE_FILES
+    ${LLAMACPP_LIB_DIR}/ggml.c
+    ${LLAMACPP_LIB_DIR}/ggml-alloc.c
+    ${LLAMACPP_LIB_DIR}/ggml-backend.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-backend-reg.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/amx/amx.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/amx/mmq.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/ggml-cpu.c
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/ggml-cpu.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/quants.c
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/traits.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/repack.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/unary-ops.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/binary-ops.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/vec.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/ops.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-opt.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-threading.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-quants.c
+    ${LLAMACPP_LIB_DIR}/gguf.cpp
+    ${LLAMACPP_LIB_DIR}/log.cpp
+    ${LLAMACPP_LIB_DIR}/llama-impl.cpp
+    ${LLAMACPP_LIB_DIR}/chat-parser.cpp
+    ${LLAMACPP_LIB_DIR}/json-partial.cpp
+    ${LLAMACPP_LIB_DIR}/regex-partial.cpp
+    # Multimodal support
+    ${LLAMACPP_LIB_DIR}/tools/mtmd/mtmd.cpp
+    ${LLAMACPP_LIB_DIR}/tools/mtmd/mtmd-audio.cpp
+    ${LLAMACPP_LIB_DIR}/tools/mtmd/clip.cpp
+    ${LLAMACPP_LIB_DIR}/tools/mtmd/mtmd-helper.cpp
+    ${LLAMACPP_LIB_DIR}/llama-grammar.cpp
+    ${LLAMACPP_LIB_DIR}/llama-sampling.cpp
+    ${LLAMACPP_LIB_DIR}/llama-vocab.cpp
+    ${LLAMACPP_LIB_DIR}/llama-adapter.cpp
+    ${LLAMACPP_LIB_DIR}/llama-chat.cpp
+    ${LLAMACPP_LIB_DIR}/llama-context.cpp
+    ${LLAMACPP_LIB_DIR}/llama-arch.cpp
+    ${LLAMACPP_LIB_DIR}/llama-batch.cpp
+    ${LLAMACPP_LIB_DIR}/llama-cparams.cpp
+    ${LLAMACPP_LIB_DIR}/llama-hparams.cpp
+    ${LLAMACPP_LIB_DIR}/llama.cpp
+    ${LLAMACPP_LIB_DIR}/llama-model.cpp
+    ${LLAMACPP_LIB_DIR}/llama-model-loader.cpp
+    ${LLAMACPP_LIB_DIR}/llama-model-saver.cpp
+    ${LLAMACPP_LIB_DIR}/llama-kv-cache.cpp
+    ${LLAMACPP_LIB_DIR}/llama-kv-cache-iswa.cpp
+    ${LLAMACPP_LIB_DIR}/llama-memory-hybrid.cpp
+    ${LLAMACPP_LIB_DIR}/llama-memory-recurrent.cpp
+    ${LLAMACPP_LIB_DIR}/llama-mmap.cpp
+    ${LLAMACPP_LIB_DIR}/llama-vocab.cpp
+    ${LLAMACPP_LIB_DIR}/llama-memory.cpp
+    ${LLAMACPP_LIB_DIR}/llama-io.cpp
+    ${LLAMACPP_LIB_DIR}/llama-graph.cpp
+    ${LLAMACPP_LIB_DIR}/sampling.cpp
+    ${LLAMACPP_LIB_DIR}/unicode-data.cpp
+    ${LLAMACPP_LIB_DIR}/unicode.cpp
+    ${LLAMACPP_LIB_DIR}/common.cpp
+    ${LLAMACPP_LIB_DIR}/chat.cpp
+    ${LLAMACPP_LIB_DIR}/json-schema-to-grammar.cpp
+    ${LLAMACPP_LIB_DIR}/nlohmann/json.hpp
+    ${LLAMACPP_LIB_DIR}/nlohmann/json_fwd.hpp
+    ${LLAMACPP_LIB_DIR}/minja/minja.hpp
+    ${LLAMACPP_LIB_DIR}/minja/chat-template.hpp
+    ${LLAMACPP_LIB_DIR}/anyascii.c
+    ${LLAMACPP_LIB_DIR}/cap-llama.cpp
+    ${LLAMACPP_LIB_DIR}/cap-completion.cpp
+    ${LLAMACPP_LIB_DIR}/cap-tts.cpp
+    ${LLAMACPP_LIB_DIR}/cap-ios-bridge.cpp
+    ${LLAMACPP_LIB_DIR}/cap-embedding.cpp
+    ${LLAMACPP_LIB_DIR}/cap-native-server.cpp
+    ${LLAMACPP_LIB_DIR}/vendor/cpp-httplib/httplib.cpp
+    ${CMAKE_SOURCE_DIR}/jni-utils.h
+    ${CMAKE_SOURCE_DIR}/jni.cpp
+)
+
+# Find Android libraries
+find_library(LOG_LIB log)
+find_library(ANDROID_LIB android)
+find_package(Threads REQUIRED)
+
+# ARM64 specific build function for real devices
+function(build_library_arm64 target_name)
+    add_library(
+        ${target_name}
+        SHARED
+        ${SOURCE_FILES}
+    )
+
+    # ARM64 specific compile options for real devices
+    target_compile_options(${target_name} PRIVATE
+        -march=armv8-a
+        -mtune=cortex-a76
+        -O3
+        -DNDEBUG
+        -DLM_GGML_USE_CPU
+        -DLM_GGML_CPU_GENERIC
+        -fno-finite-math-only
+        -funroll-loops
+    )
+
+    # Link with Android libraries
+    target_link_libraries(${target_name} 
+        ${LOG_LIB}
+        ${ANDROID_LIB}
+        Threads::Threads
+    )
+    
+    # Set output name and directory
+    set_target_properties(${target_name} PROPERTIES
+        OUTPUT_NAME "llama-cpp-arm64"
+        LIBRARY_OUTPUT_DIRECTORY "${CMAKE_CURRENT_SOURCE_DIR}/jniLibs/arm64-v8a"
+    )
+endfunction()
+
+# Build ARM64 library for real devices
+build_library_arm64(llama-cpp-arm64)
+
+# Print build information
+message(STATUS "Building llama-cpp for Android ARM64 (real devices)")
+message(STATUS "Source directory: ${LLAMACPP_LIB_DIR}")
+message(STATUS "Architecture: ARM64 (arm64-v8a)")
+message(STATUS "Optimizations: ARMv8-A, Cortex-A76 tuning")
+message(STATUS "Output directory: ${CMAKE_CURRENT_SOURCE_DIR}/jniLibs/arm64-v8a")

package/android/src/main/CMakeLists-x86_64.txt

@@ -0,0 +1,141 @@
+cmake_minimum_required(VERSION 3.10)
+
+project(llama-cpp)
+
+set(CMAKE_CXX_STANDARD 17)
+set(LLAMACPP_LIB_DIR ${CMAKE_SOURCE_DIR}/../../../cpp)
+
+include_directories(
+    ${LLAMACPP_LIB_DIR}
+    ${LLAMACPP_LIB_DIR}/ggml-cpu
+    ${LLAMACPP_LIB_DIR}/tools/mtmd
+    ${LLAMACPP_LIB_DIR}/vendor/cpp-httplib
+)
+
+set(
+    SOURCE_FILES
+    ${LLAMACPP_LIB_DIR}/ggml.c
+    ${LLAMACPP_LIB_DIR}/ggml-alloc.c
+    ${LLAMACPP_LIB_DIR}/ggml-backend.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-backend-reg.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/amx/amx.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/amx/mmq.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/ggml-cpu.c
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/ggml-cpu.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/quants.c
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/traits.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/repack.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/unary-ops.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/binary-ops.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/vec.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/ops.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-opt.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-threading.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-quants.c
+    ${LLAMACPP_LIB_DIR}/gguf.cpp
+    ${LLAMACPP_LIB_DIR}/log.cpp
+    ${LLAMACPP_LIB_DIR}/llama-impl.cpp
+    ${LLAMACPP_LIB_DIR}/chat-parser.cpp
+    ${LLAMACPP_LIB_DIR}/json-partial.cpp
+    ${LLAMACPP_LIB_DIR}/regex-partial.cpp
+    # Multimodal support
+    ${LLAMACPP_LIB_DIR}/tools/mtmd/mtmd.cpp
+    ${LLAMACPP_LIB_DIR}/tools/mtmd/mtmd-audio.cpp
+    ${LLAMACPP_LIB_DIR}/tools/mtmd/clip.cpp
+    ${LLAMACPP_LIB_DIR}/tools/mtmd/mtmd-helper.cpp
+    ${LLAMACPP_LIB_DIR}/llama-grammar.cpp
+    ${LLAMACPP_LIB_DIR}/llama-sampling.cpp
+    ${LLAMACPP_LIB_DIR}/llama-vocab.cpp
+    ${LLAMACPP_LIB_DIR}/llama-adapter.cpp
+    ${LLAMACPP_LIB_DIR}/llama-chat.cpp
+    ${LLAMACPP_LIB_DIR}/llama-context.cpp
+    ${LLAMACPP_LIB_DIR}/llama-arch.cpp
+    ${LLAMACPP_LIB_DIR}/llama-batch.cpp
+    ${LLAMACPP_LIB_DIR}/llama-cparams.cpp
+    ${LLAMACPP_LIB_DIR}/llama-hparams.cpp
+    ${LLAMACPP_LIB_DIR}/llama.cpp
+    ${LLAMACPP_LIB_DIR}/llama-model.cpp
+    ${LLAMACPP_LIB_DIR}/llama-model-loader.cpp
+    ${LLAMACPP_LIB_DIR}/llama-model-saver.cpp
+    ${LLAMACPP_LIB_DIR}/llama-kv-cache.cpp
+    ${LLAMACPP_LIB_DIR}/llama-kv-cache-iswa.cpp
+    ${LLAMACPP_LIB_DIR}/llama-memory-hybrid.cpp
+    ${LLAMACPP_LIB_DIR}/llama-memory-recurrent.cpp
+    ${LLAMACPP_LIB_DIR}/llama-mmap.cpp
+    ${LLAMACPP_LIB_DIR}/llama-vocab.cpp
+    ${LLAMACPP_LIB_DIR}/llama-memory.cpp
+    ${LLAMACPP_LIB_DIR}/llama-io.cpp
+    ${LLAMACPP_LIB_DIR}/llama-graph.cpp
+    ${LLAMACPP_LIB_DIR}/sampling.cpp
+    ${LLAMACPP_LIB_DIR}/unicode-data.cpp
+    ${LLAMACPP_LIB_DIR}/unicode.cpp
+    ${LLAMACPP_LIB_DIR}/common.cpp
+    ${LLAMACPP_LIB_DIR}/chat.cpp
+    ${LLAMACPP_LIB_DIR}/json-schema-to-grammar.cpp
+    ${LLAMACPP_LIB_DIR}/nlohmann/json.hpp
+    ${LLAMACPP_LIB_DIR}/nlohmann/json_fwd.hpp
+    ${LLAMACPP_LIB_DIR}/minja/minja.hpp
+    ${LLAMACPP_LIB_DIR}/minja/chat-template.hpp
+    ${LLAMACPP_LIB_DIR}/anyascii.c
+    ${LLAMACPP_LIB_DIR}/cap-llama.cpp
+    ${LLAMACPP_LIB_DIR}/cap-completion.cpp
+    ${LLAMACPP_LIB_DIR}/cap-tts.cpp
+    ${LLAMACPP_LIB_DIR}/cap-ios-bridge.cpp
+    ${LLAMACPP_LIB_DIR}/cap-embedding.cpp
+    ${LLAMACPP_LIB_DIR}/cap-native-server.cpp
+    ${LLAMACPP_LIB_DIR}/vendor/cpp-httplib/httplib.cpp
+    ${CMAKE_SOURCE_DIR}/jni-utils.h
+    ${CMAKE_SOURCE_DIR}/jni.cpp
+)
+
+find_library(LOG_LIB log)
+find_package(Threads REQUIRED)
+
+# X86_64 specific build function for emulator
+function(build_library_x86_64 target_name)
+    add_library(
+        ${target_name}
+        SHARED
+        ${SOURCE_FILES}
+    )
+
+    # X86_64 specific compile options for emulator
+    target_compile_options(${target_name} PRIVATE
+        -march=x86-64
+        -mtune=generic
+        -mavx2
+        -mavx
+        -msse3
+        -msse
+        -mfma
+        -mf16c
+        -O3
+        -DNDEBUG
+        -DLM_GGML_USE_CPU
+        -DLM_GGML_CPU_GENERIC
+        -DLM_GGML_USE_AVX2
+        -DLM_GGML_USE_AVX
+        -DLM_GGML_USE_SSE3
+        -DLM_GGML_USE_SSE
+        -DLM_GGML_USE_FMA
+        -DLM_GGML_USE_F16C
+    )
+
+    target_link_libraries(${target_name} ${LOG_LIB} Threads::Threads)
+    
+    # Set output name
+    set_target_properties(${target_name} PROPERTIES
+        OUTPUT_NAME "llama-cpp-x86_64"
+        LIBRARY_OUTPUT_DIRECTORY "${CMAKE_CURRENT_SOURCE_DIR}/jniLibs/x86_64"
+    )
+endfunction()
+
+# Build x86_64 library for emulator
+build_library_x86_64(llama-cpp-x86_64)
+
+# Print build information
+message(STATUS "Building llama-cpp for Android x86_64 (emulator)")
+message(STATUS "Source directory: ${LLAMACPP_LIB_DIR}")
+message(STATUS "Architecture: x86_64")
+message(STATUS "Optimizations: AVX2, AVX, SSE3, SSE, FMA, F16C")
+message(STATUS "Output directory: ${CMAKE_CURRENT_SOURCE_DIR}/jniLibs/x86_64")

package/android/src/main/CMakeLists.txt

@@ -0,0 +1,138 @@
+cmake_minimum_required(VERSION 3.10)
+
+project(llama-cpp)
+
+set(CMAKE_CXX_STANDARD 17)
+set(LLAMACPP_LIB_DIR ${CMAKE_SOURCE_DIR}/../../../cpp)
+
+include_directories(
+    ${LLAMACPP_LIB_DIR}
+    ${LLAMACPP_LIB_DIR}/ggml-cpu
+    ${LLAMACPP_LIB_DIR}/tools/mtmd
+    ${LLAMACPP_LIB_DIR}/vendor/cpp-httplib
+)
+
+set(
+    SOURCE_FILES
+    ${LLAMACPP_LIB_DIR}/ggml.c
+    ${LLAMACPP_LIB_DIR}/ggml-alloc.c
+    ${LLAMACPP_LIB_DIR}/ggml-backend.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-backend-reg.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/amx/amx.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/amx/mmq.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/ggml-cpu.c
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/ggml-cpu.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/quants.c
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/traits.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/repack.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/unary-ops.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/binary-ops.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/vec.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-cpu/ops.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-opt.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-threading.cpp
+    ${LLAMACPP_LIB_DIR}/ggml-quants.c
+    ${LLAMACPP_LIB_DIR}/gguf.cpp
+    ${LLAMACPP_LIB_DIR}/log.cpp
+    ${LLAMACPP_LIB_DIR}/llama-impl.cpp
+    ${LLAMACPP_LIB_DIR}/chat-parser.cpp
+    ${LLAMACPP_LIB_DIR}/json-partial.cpp
+    ${LLAMACPP_LIB_DIR}/regex-partial.cpp
+    # Multimodal support
+    ${LLAMACPP_LIB_DIR}/tools/mtmd/mtmd.cpp
+    ${LLAMACPP_LIB_DIR}/tools/mtmd/mtmd-audio.cpp
+    ${LLAMACPP_LIB_DIR}/tools/mtmd/clip.cpp
+    ${LLAMACPP_LIB_DIR}/tools/mtmd/mtmd-helper.cpp
+    ${LLAMACPP_LIB_DIR}/llama-grammar.cpp
+    ${LLAMACPP_LIB_DIR}/llama-sampling.cpp
+    ${LLAMACPP_LIB_DIR}/llama-vocab.cpp
+    ${LLAMACPP_LIB_DIR}/llama-adapter.cpp
+    ${LLAMACPP_LIB_DIR}/llama-chat.cpp
+    ${LLAMACPP_LIB_DIR}/llama-context.cpp
+    ${LLAMACPP_LIB_DIR}/llama-arch.cpp
+    ${LLAMACPP_LIB_DIR}/llama-batch.cpp
+    ${LLAMACPP_LIB_DIR}/llama-cparams.cpp
+    ${LLAMACPP_LIB_DIR}/llama-hparams.cpp
+    ${LLAMACPP_LIB_DIR}/llama.cpp
+    ${LLAMACPP_LIB_DIR}/llama-model.cpp
+    ${LLAMACPP_LIB_DIR}/llama-model-loader.cpp
+    ${LLAMACPP_LIB_DIR}/llama-model-saver.cpp
+    ${LLAMACPP_LIB_DIR}/llama-kv-cache.cpp
+    ${LLAMACPP_LIB_DIR}/llama-kv-cache-iswa.cpp
+    ${LLAMACPP_LIB_DIR}/llama-memory-hybrid.cpp
+    ${LLAMACPP_LIB_DIR}/llama-memory-recurrent.cpp
+    ${LLAMACPP_LIB_DIR}/llama-mmap.cpp
+    ${LLAMACPP_LIB_DIR}/llama-vocab.cpp
+    ${LLAMACPP_LIB_DIR}/llama-memory.cpp
+    ${LLAMACPP_LIB_DIR}/llama-io.cpp
+    ${LLAMACPP_LIB_DIR}/llama-graph.cpp
+    ${LLAMACPP_LIB_DIR}/sampling.cpp
+    ${LLAMACPP_LIB_DIR}/unicode-data.cpp
+    ${LLAMACPP_LIB_DIR}/unicode.cpp
+    ${LLAMACPP_LIB_DIR}/common.cpp
+    ${LLAMACPP_LIB_DIR}/chat.cpp
+    ${LLAMACPP_LIB_DIR}/json-schema-to-grammar.cpp
+    ${LLAMACPP_LIB_DIR}/nlohmann/json.hpp
+    ${LLAMACPP_LIB_DIR}/nlohmann/json_fwd.hpp
+    ${LLAMACPP_LIB_DIR}/minja/minja.hpp
+    ${LLAMACPP_LIB_DIR}/minja/chat-template.hpp
+    ${LLAMACPP_LIB_DIR}/anyascii.c
+    ${LLAMACPP_LIB_DIR}/cap-llama.cpp
+    ${LLAMACPP_LIB_DIR}/cap-completion.cpp
+    ${LLAMACPP_LIB_DIR}/cap-tts.cpp
+    ${LLAMACPP_LIB_DIR}/cap-ios-bridge.cpp
+    ${LLAMACPP_LIB_DIR}/cap-embedding.cpp
+    ${LLAMACPP_LIB_DIR}/cap-native-server.cpp
+    ${LLAMACPP_LIB_DIR}/vendor/cpp-httplib/httplib.cpp
+    ${CMAKE_SOURCE_DIR}/jni-utils.h
+    ${CMAKE_SOURCE_DIR}/jni.cpp
+)
+
+# Find Android libraries
+find_library(LOG_LIB log)
+find_library(ANDROID_LIB android)
+find_package(Threads REQUIRED)
+
+# ARM64 specific build function for real devices
+function(build_library_arm64 target_name)
+    add_library(
+        ${target_name}
+        SHARED
+        ${SOURCE_FILES}
+    )
+
+    # ARM64 specific compile options for real devices
+    target_compile_options(${target_name} PRIVATE
+        -march=armv8-a
+        -mtune=cortex-a76
+        -O3
+        -DNDEBUG
+        -DLM_GGML_USE_CPU
+        -DLM_GGML_CPU_GENERIC
+        -fno-finite-math-only
+        -funroll-loops
+    )
+
+    # Link with Android libraries
+    target_link_libraries(${target_name} 
+        ${LOG_LIB}
+        ${ANDROID_LIB}
+        Threads::Threads
+    )
+    
+    # Set output name and directory
+    set_target_properties(${target_name} PROPERTIES
+        OUTPUT_NAME "llama-cpp-arm64"
+        LIBRARY_OUTPUT_DIRECTORY "${CMAKE_CURRENT_SOURCE_DIR}/jniLibs/arm64-v8a"
+    )
+endfunction()
+
+# Build ARM64 library for real devices
+build_library_arm64(llama-cpp-arm64)
+
+# Print build information
+message(STATUS "Building llama-cpp for Android ARM64 (real devices)")
+message(STATUS "Source directory: ${LLAMACPP_LIB_DIR}")
+message(STATUS "Architecture: ARM64 (arm64-v8a)")
+message(STATUS "Optimizations: ARMv8-A, Cortex-A76 tuning")
+message(STATUS "Output directory: ${CMAKE_CURRENT_SOURCE_DIR}/jniLibs/arm64-v8a")