@fugood/llama.node 0.2.3 → 0.3.1

package/src/llama.cpp/examples/server/tests/requirements.txt

@@ -1,6 +1,6 @@
 aiohttp~=3.9.3
 behave~=1.2.6
 huggingface_hub~=0.20.3
-numpy~=1.24.4
-openai~=0.25.0
+numpy~=1.26.4
+openai~=1.30.3
 prometheus-client~=0.20.0

package/src/llama.cpp/examples/server/utils.hpp

@@ -116,45 +116,37 @@ static inline void server_log(const char * level, const char * function, int lin
 // chat template utils
 //
 
-// Check if the template supplied via "--chat-template" is supported or not. Returns true if it's valid
-inline bool verify_custom_template(const std::string & tmpl) {
-    llama_chat_message chat[] = {{"user", "test"}};
-    int res = llama_chat_apply_template(nullptr, tmpl.c_str(), chat, 1, true, nullptr, 0);
-    return res >= 0;
-}
-
 // Format given chat. If tmpl is empty, we take the template from model metadata
 inline std::string format_chat(const struct llama_model * model, const std::string & tmpl, const std::vector<json> & messages) {
-    size_t alloc_size = 0;
-    // vector holding all allocated string to be passed to llama_chat_apply_template
-    std::vector<std::string> str(messages.size() * 2);
-    std::vector<llama_chat_message> chat(messages.size());
+    std::vector<llama_chat_msg> chat;
 
     for (size_t i = 0; i < messages.size(); ++i) {
         const auto & curr_msg = messages[i];
-        str[i*2 + 0]    = json_value(curr_msg, "role",    std::string(""));
-        str[i*2 + 1]    = json_value(curr_msg, "content", std::string(""));
-        alloc_size     += str[i*2 + 1].length();
-        chat[i].role    = str[i*2 + 0].c_str();
-        chat[i].content = str[i*2 + 1].c_str();
-    }
 
-    const char * ptr_tmpl = tmpl.empty() ? nullptr : tmpl.c_str();
-    std::vector<char> buf(alloc_size * 2);
-
-    // run the first time to get the total output length
-    int32_t res = llama_chat_apply_template(model, ptr_tmpl, chat.data(), chat.size(), true, buf.data(), buf.size());
+        std::string role = json_value(curr_msg, "role", std::string(""));
+
+        std::string content;
+        if (curr_msg.contains("content")) {
+            if (curr_msg["content"].is_string()) {
+                content = curr_msg["content"].get<std::string>();
+            } else if (curr_msg["content"].is_array()) {
+                for (const auto & part : curr_msg["content"]) {
+                    if (part.contains("text")) {
+                        content += "\n" + part["text"].get<std::string>();
+                    }
+                }
+            } else {
+                throw std::runtime_error("Invalid 'content' type (ref: https://github.com/ggerganov/llama.cpp/issues/8367)");
+            }
+        } else {
+            throw std::runtime_error("Missing 'content' (ref: https://github.com/ggerganov/llama.cpp/issues/8367)");
+        }
 
-    // if it turns out that our buffer is too small, we resize it
-    if ((size_t) res > buf.size()) {
-        buf.resize(res);
-        res = llama_chat_apply_template(model, ptr_tmpl, chat.data(), chat.size(), true, buf.data(), buf.size());
+        chat.push_back({role, content});
     }
 
-    const std::string formatted_chat(buf.data(), res);
-
+    auto formatted_chat = llama_chat_apply_template(model, tmpl, chat, true);
     LOG_VERBOSE("formatted_chat", {{"text", formatted_chat.c_str()}});
-
     return formatted_chat;
 }
 
@@ -260,6 +252,13 @@ static size_t common_part(const std::vector<llama_token> & a, const std::vector<
     return i;
 }
 
+static size_t common_part(const std::string & a, const std::string & b) {
+    size_t i;
+    for (i = 0; i < a.size() && i < b.size() && a[i] == b[i]; i++) {}
+
+    return i;
+}
+
 static bool ends_with(const std::string & str, const std::string & suffix) {
     return str.size() >= suffix.size() && 0 == str.compare(str.size() - suffix.size(), suffix.size(), suffix);
 }

package/src/llama.cpp/examples/simple/CMakeLists.txt

@@ -1,4 +1,4 @@
-set(TARGET simple)
+set(TARGET llama-simple)
 add_executable(${TARGET} simple.cpp)
 install(TARGETS ${TARGET} RUNTIME)
 target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})

package/src/llama.cpp/examples/simple/simple.cpp

@@ -6,28 +6,27 @@
 #include <string>
 #include <vector>
 
-int main(int argc, char ** argv) {
-    gpt_params params;
+static void print_usage(int argc, char ** argv, const gpt_params & params) {
+    gpt_params_print_usage(argc, argv, params);
 
-    if (argc == 1 || argv[1][0] == '-') {
-        printf("usage: %s MODEL_PATH [PROMPT]\n" , argv[0]);
-        return 1 ;
-    }
+    LOG_TEE("\nexample usage:\n");
+    LOG_TEE("\n    %s -m model.gguf -p \"Hello my name is\" -n 32\n", argv[0]);
+    LOG_TEE("\n");
+}
 
-    if (argc >= 2) {
-        params.model = argv[1];
-    }
+int main(int argc, char ** argv) {
+    gpt_params params;
 
-    if (argc >= 3) {
-        params.prompt = argv[2];
-    }
+    params.prompt = "Hello my name is";
+    params.n_predict = 32;
 
-    if (params.prompt.empty()) {
-        params.prompt = "Hello my name is";
+    if (!gpt_params_parse(argc, argv, params)) {
+        print_usage(argc, argv, params);
+        return 1;
     }
 
     // total length of the sequence including the prompt
-    const int n_len = 32;
+    const int n_predict = params.n_predict;
 
     // init LLM
 
@@ -36,9 +35,7 @@ int main(int argc, char ** argv) {
 
     // initialize the model
 
-    llama_model_params model_params = llama_model_default_params();
-
-    // model_params.n_gpu_layers = 99; // offload all layers to the GPU
+    llama_model_params model_params = llama_model_params_from_gpt_params(params);
 
     llama_model * model = llama_load_model_from_file(params.model.c_str(), model_params);
 
@@ -49,12 +46,7 @@ int main(int argc, char ** argv) {
 
     // initialize the context
 
-    llama_context_params ctx_params = llama_context_default_params();
-
-    ctx_params.seed  = 1234;
-    ctx_params.n_ctx = 2048;
-    ctx_params.n_threads = params.n_threads;
-    ctx_params.n_threads_batch = params.n_threads_batch == -1 ? params.n_threads : params.n_threads_batch;
+    llama_context_params ctx_params = llama_context_params_from_gpt_params(params);
 
     llama_context * ctx = llama_new_context_with_model(model, ctx_params);
 
@@ -69,14 +61,14 @@ int main(int argc, char ** argv) {
     tokens_list = ::llama_tokenize(ctx, params.prompt, true);
 
     const int n_ctx    = llama_n_ctx(ctx);
-    const int n_kv_req = tokens_list.size() + (n_len - tokens_list.size());
+    const int n_kv_req = tokens_list.size() + (n_predict - tokens_list.size());
 
-    LOG_TEE("\n%s: n_len = %d, n_ctx = %d, n_kv_req = %d\n", __func__, n_len, n_ctx, n_kv_req);
+    LOG_TEE("\n%s: n_predict = %d, n_ctx = %d, n_kv_req = %d\n", __func__, n_predict, n_ctx, n_kv_req);
 
     // make sure the KV cache is big enough to hold all the prompt and generated tokens
     if (n_kv_req > n_ctx) {
         LOG_TEE("%s: error: n_kv_req > n_ctx, the required KV cache size is not big enough\n", __func__);
-        LOG_TEE("%s:        either reduce n_len or increase n_ctx\n", __func__);
+        LOG_TEE("%s:        either reduce n_predict or increase n_ctx\n", __func__);
         return 1;
     }
 
@@ -115,7 +107,7 @@ int main(int argc, char ** argv) {
 
     const auto t_main_start = ggml_time_us();
 
-    while (n_cur <= n_len) {
+    while (n_cur <= n_predict) {
         // sample the next token
         {
             auto   n_vocab = llama_n_vocab(model);
@@ -134,7 +126,7 @@ int main(int argc, char ** argv) {
             const llama_token new_token_id = llama_sample_token_greedy(ctx, &candidates_p);
 
             // is it an end of generation?
-            if (llama_token_is_eog(model, new_token_id) || n_cur == n_len) {
+            if (llama_token_is_eog(model, new_token_id) || n_cur == n_predict) {
                 LOG_TEE("\n");
 
                 break;

package/src/llama.cpp/examples/speculative/CMakeLists.txt

@@ -1,4 +1,4 @@
-set(TARGET speculative)
+set(TARGET llama-speculative)
 add_executable(${TARGET} speculative.cpp)
 install(TARGETS ${TARGET} RUNTIME)
 target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})

package/src/llama.cpp/examples/speculative/speculative.cpp

@@ -27,7 +27,8 @@ struct seq_draft {
 int main(int argc, char ** argv) {
     gpt_params params;
 
-    if (gpt_params_parse(argc, argv, params) == false) {
+    if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
 

package/src/llama.cpp/examples/sycl/CMakeLists.txt

@@ -2,7 +2,7 @@
 #  Copyright (C) 2024 Intel Corporation
 #  SPDX-License-Identifier: MIT
 
-set(TARGET ls-sycl-device)
+set(TARGET llama-ls-sycl-device)
 add_executable(${TARGET} ls-sycl-device.cpp)
 install(TARGETS ${TARGET} RUNTIME)
 target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})

package/src/llama.cpp/examples/sycl/build.sh

@@ -0,0 +1,23 @@
+
+#  MIT license
+#  Copyright (C) 2024 Intel Corporation
+#  SPDX-License-Identifier: MIT
+
+mkdir -p build
+cd build
+source /opt/intel/oneapi/setvars.sh
+
+#for FP16
+#cmake .. -DGGML_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DGGML_SYCL_F16=ON # faster for long-prompt inference
+
+#for FP32
+cmake .. -DGGML_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
+
+#build example/main
+#cmake --build . --config Release --target main
+
+#build example/llama-bench
+#cmake --build . --config Release --target llama-bench
+
+#build all binary
+cmake --build . --config Release -j -v

package/src/llama.cpp/examples/sycl/run-llama2.sh

@@ -0,0 +1,36 @@
+#!/bin/bash
+
+#  MIT license
+#  Copyright (C) 2024 Intel Corporation
+#  SPDX-License-Identifier: MIT
+
+INPUT2="Building a website can be done in 10 simple steps:\nStep 1:"
+source /opt/intel/oneapi/setvars.sh
+
+if [ $# -gt 0 ]; then
+    GGML_SYCL_DEVICE=$1
+    GGML_SYCL_SINGLE_GPU=1
+else
+    GGML_SYCL_DEVICE=0
+    GGML_SYCL_SINGLE_GPU=0
+fi
+
+#export GGML_SYCL_DEBUG=1
+
+
+#ZES_ENABLE_SYSMAN=1, Support to get free memory of GPU by sycl::aspect::ext_intel_free_memory. Recommended to use when --split-mode = layer.
+
+if [ $GGML_SYCL_SINGLE_GPU -eq 1 ]; then
+    echo "use $GGML_SYCL_DEVICE as main GPU"
+    #use signle GPU only
+    ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 400 -e -ngl 33 -s 0 -mg $GGML_SYCL_DEVICE -sm none
+else
+    #use multiple GPUs with same max compute units
+    ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 400 -e -ngl 33 -s 0
+fi
+
+#use main GPU only
+#ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 400 -e -ngl 33 -s 0 -mg $GGML_SYCL_DEVICE -sm none
+
+#use multiple GPUs with same max compute units
+#ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 400 -e -ngl 33 -s 0

package/src/llama.cpp/examples/sycl/win-build-sycl.bat

@@ -0,0 +1,33 @@
+
+::  MIT license
+::  Copyright (C) 2024 Intel Corporation
+::  SPDX-License-Identifier: MIT
+
+
+IF not exist build (mkdir build)
+cd build
+if %errorlevel% neq 0 goto ERROR
+
+@call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64 --force
+if %errorlevel% neq 0 goto ERROR
+
+::  for FP16
+::  faster for long-prompt inference
+::  cmake -G "MinGW Makefiles" ..  -DGGML_SYCL=ON -DCMAKE_CXX_COMPILER=icx -DBUILD_SHARED_LIBS=ON -DCMAKE_BUILD_TYPE=Release -DGGML_SYCL_F16=ON
+
+::  for FP32
+cmake -G "Ninja" ..  -DGGML_SYCL=ON -DCMAKE_C_COMPILER=cl -DCMAKE_CXX_COMPILER=icx -DBUILD_SHARED_LIBS=ON -DCMAKE_BUILD_TYPE=Release
+if %errorlevel% neq 0 goto ERROR
+::  build example/main only
+::  make main
+
+::  build all binary
+cmake --build . -j
+if %errorlevel% neq 0 goto ERROR
+
+cd ..
+exit /B 0
+
+:ERROR
+echo comomand error: %errorlevel%
+exit /B %errorlevel%

package/src/llama.cpp/examples/sycl/win-run-llama2.bat

@@ -0,0 +1,9 @@
+::  MIT license
+::  Copyright (C) 2024 Intel Corporation
+::  SPDX-License-Identifier: MIT
+
+set INPUT2="Building a website can be done in 10 simple steps:\nStep 1:"
+@call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64 --force
+
+
+.\build\bin\main.exe -m models\llama-2-7b.Q4_0.gguf -p %INPUT2% -n 400 -e -ngl 33 -s 0

package/src/llama.cpp/examples/tokenize/CMakeLists.txt

@@ -1,4 +1,4 @@
-set(TARGET tokenize)
+set(TARGET llama-tokenize)
 add_executable(${TARGET} tokenize.cpp)
 install(TARGETS ${TARGET} RUNTIME)
 target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})

package/src/llama.cpp/examples/tokenize/tokenize.cpp

@@ -29,7 +29,9 @@ static void print_usage_information(const char * argv0, FILE * stream) {
     fprintf(stream, "    -p PROMPT, --prompt PROMPT           read prompt from the argument.\n");
     fprintf(stream, "    --stdin                              read prompt from standard input.\n");
     fprintf(stream, "    --no-bos                             do not ever add a BOS token to the prompt, even if normally the model uses a BOS token.\n");
+    fprintf(stream, "    --no-parse-special                   do not parse control tokens.\n");
     fprintf(stream, "    --log-disable                        disable logs. Makes stderr quiet when loading the model.\n");
+    fprintf(stream, "    --show-count                         print the total number of tokens.\n");
 }
 
 static void llama_log_callback_null(ggml_log_level level, const char * text, void * user_data) {
@@ -161,7 +163,7 @@ static void write_utf8_cstr_to_stdout(const char * str, bool & invalid_utf8) {
                 printf(">");
                 return;
             }
-            GGML_ASSERT(false && "MultiByteToWideChar() failed in an unexpected way.");
+            GGML_ABORT("MultiByteToWideChar() failed in an unexpected way.");
         }
 
         LPWSTR wstr = (LPWSTR) calloc(length_needed+1, sizeof(*wstr));
@@ -194,7 +196,9 @@ int main(int raw_argc, char ** raw_argv) {
     // variables where to put any arguments we see.
     bool printing_ids = false;
     bool no_bos = false;
+    bool no_parse_special = false;
     bool disable_logging = false;
+    bool show_token_count = false;
     const char * model_path = NULL;
     const char * prompt_path = NULL;
     const char * prompt_arg = NULL;
@@ -227,6 +231,9 @@ int main(int raw_argc, char ** raw_argv) {
         else if (arg == "--no-bos") {
             no_bos = true;
         }
+        else if (arg == "--no-parse-special") {
+            no_parse_special = true;
+        }
         else if (arg == "-p" || arg == "--prompt") {
             if (prompt_set) {
                 fprintf(stderr, "Error: -p or --prompt specified multiple times.\n");
@@ -249,6 +256,9 @@ int main(int raw_argc, char ** raw_argv) {
         else if (arg == "--log-disable") {
             disable_logging = true;
         }
+        else if (arg == "--show-count") {
+            show_token_count = true;
+        }
         else {
             fprintf(stderr, "Error: unknown option '%s'\n", argv[iarg].c_str());
             return 1;
@@ -354,9 +364,10 @@ int main(int raw_argc, char ** raw_argv) {
 
     const bool model_wants_add_bos = llama_should_add_bos_token(model);
     const bool add_bos = model_wants_add_bos && !no_bos;
+    const bool parse_special = !no_parse_special;
 
     std::vector<llama_token> tokens;
-    tokens = ::llama_tokenize(model, prompt, add_bos, true);
+    tokens = ::llama_tokenize(model, prompt, add_bos, parse_special);
 
     if (printing_ids) {
         printf("[");
@@ -384,6 +395,9 @@ int main(int raw_argc, char ** raw_argv) {
         printf("]\n");
     }
 
+    if (show_token_count) {
+        printf("Total number of tokens: %ld\n", tokens.size());
+    }
     // silence valgrind
     llama_free(ctx);
     llama_free_model(model);

package/src/llama.cpp/ggml/CMakeLists.txt

@@ -0,0 +1,253 @@
+cmake_minimum_required(VERSION 3.14) # for add_link_options and implicit target directories.
+project("ggml" C CXX)
+include(CheckIncludeFileCXX)
+
+set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
+
+if (NOT XCODE AND NOT MSVC AND NOT CMAKE_BUILD_TYPE)
+    set(CMAKE_BUILD_TYPE Release CACHE STRING "Build type" FORCE)
+    set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Debug" "Release" "MinSizeRel" "RelWithDebInfo")
+endif()
+
+if (CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR)
+    set(GGML_STANDALONE ON)
+
+    set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
+
+    # configure project version
+    # TODO
+else()
+    set(GGML_STANDALONE OFF)
+endif()
+
+if (EMSCRIPTEN)
+    set(BUILD_SHARED_LIBS_DEFAULT OFF)
+
+    option(GGML_WASM_SINGLE_FILE "ggml: embed WASM inside the generated ggml.js" ON)
+else()
+    if (MINGW)
+        set(BUILD_SHARED_LIBS_DEFAULT OFF)
+    else()
+        set(BUILD_SHARED_LIBS_DEFAULT ON)
+    endif()
+endif()
+
+option(BUILD_SHARED_LIBS "ggml: build shared libraries" ${BUILD_SHARED_LIBS_DEFAULT})
+
+#
+# option list
+#
+
+# TODO: mark all options as advanced when not GGML_STANDALONE
+
+if (APPLE)
+    set(GGML_METAL_DEFAULT ON)
+    set(GGML_BLAS_DEFAULT ON)
+    set(GGML_BLAS_VENDOR_DEFAULT "Apple")
+else()
+    set(GGML_METAL_DEFAULT OFF)
+    set(GGML_BLAS_DEFAULT OFF)
+    set(GGML_BLAS_VENDOR_DEFAULT "Generic")
+endif()
+
+if (CMAKE_CROSSCOMPILING)
+    set(GGML_NATIVE_DEFAULT OFF)
+else()
+    set(GGML_NATIVE_DEFAULT ON)
+endif()
+
+# general
+option(GGML_STATIC "ggml: static link libraries"         OFF)
+option(GGML_NATIVE "ggml: enable -march=native flag"     ${GGML_NATIVE_DEFAULT})
+option(GGML_LTO    "ggml: enable link time optimization" OFF)
+option(GGML_CCACHE "ggml: use ccache if available"       ON)
+
+# debug
+option(GGML_ALL_WARNINGS           "ggml: enable all compiler warnings"                   ON)
+option(GGML_ALL_WARNINGS_3RD_PARTY "ggml: enable all compiler warnings in 3rd party libs" OFF)
+option(GGML_GPROF                  "ggml: enable gprof"                                   OFF)
+
+# build
+option(GGML_FATAL_WARNINGS    "ggml: enable -Werror flag"    OFF)
+
+# sanitizers
+option(GGML_SANITIZE_THREAD    "ggml: enable thread sanitizer"    OFF)
+option(GGML_SANITIZE_ADDRESS   "ggml: enable address sanitizer"   OFF)
+option(GGML_SANITIZE_UNDEFINED "ggml: enable undefined sanitizer" OFF)
+
+# instruction set specific
+if (GGML_NATIVE OR NOT GGML_NATIVE_DEFAULT)
+    set(INS_ENB OFF)
+else()
+    set(INS_ENB ON)
+endif()
+
+option(GGML_CPU_HBM     "ggml: use memkind for CPU HBM" OFF)
+
+option(GGML_AVX         "ggml: enable AVX"              ${INS_ENB})
+option(GGML_AVX2        "ggml: enable AVX2"             ${INS_ENB})
+option(GGML_AVX512      "ggml: enable AVX512"           OFF)
+option(GGML_AVX512_VBMI "ggml: enable AVX512-VBMI"      OFF)
+option(GGML_AVX512_VNNI "ggml: enable AVX512-VNNI"      OFF)
+option(GGML_AVX512_BF16 "ggml: enable AVX512-BF16"      OFF)
+option(GGML_FMA         "ggml: enable FMA"              ${INS_ENB})
+if (NOT MSVC)
+    option(GGML_F16C    "ggml: enable F16C"             ${INS_ENB}) # in MSVC F16C is implied with AVX2/AVX512
+endif()
+option(GGML_LASX        "ggml: enable lasx"             ON)
+option(GGML_LSX         "ggml: enable lsx"              ON)
+option(GGML_SVE         "ggml: enable SVE"              OFF)
+
+if (WIN32)
+    set(GGML_WIN_VER "0x602" CACHE STRING "ggml: Windows Version")
+endif()
+
+# ggml core
+set(GGML_SCHED_MAX_COPIES  "4" CACHE STRING "ggml: max input copies for pipeline parallelism")
+
+# 3rd party libs / backends
+option(GGML_ACCELERATE                      "ggml: enable Accelerate framework"               ON)
+option(GGML_BLAS                            "ggml: use BLAS"                                  ${GGML_BLAS_DEFAULT})
+set(GGML_BLAS_VENDOR ${GGML_BLAS_VENDOR_DEFAULT} CACHE STRING
+                                            "ggml: BLAS library vendor")
+option(GGML_LLAMAFILE                       "ggml: use LLAMAFILE"                             OFF)
+
+option(GGML_CUDA                            "ggml: use CUDA"                                  OFF)
+option(GGML_MUSA                            "ggml: use MUSA"                                  OFF)
+option(GGML_CUDA_FORCE_DMMV                 "ggml: use dmmv instead of mmvq CUDA kernels"     OFF)
+option(GGML_CUDA_FORCE_MMQ                  "ggml: use mmq kernels instead of cuBLAS"         OFF)
+option(GGML_CUDA_FORCE_CUBLAS               "ggml: always use cuBLAS instead of mmq kernels"  OFF)
+set   (GGML_CUDA_DMMV_X   "32" CACHE STRING "ggml: x stride for dmmv CUDA kernels")
+set   (GGML_CUDA_MMV_Y     "1" CACHE STRING "ggml: y block size for mmv CUDA kernels")
+option(GGML_CUDA_F16                        "ggml: use 16 bit floats for some calculations"   OFF)
+set   (GGML_CUDA_KQUANTS_ITER "2" CACHE STRING
+                                            "ggml: iters./thread per block for Q2_K/Q6_K")
+set   (GGML_CUDA_PEER_MAX_BATCH_SIZE "128" CACHE STRING
+                                            "ggml: max. batch size for using peer access")
+option(GGML_CUDA_NO_PEER_COPY               "ggml: do not use peer to peer copies"            OFF)
+option(GGML_CUDA_NO_VMM                     "ggml: do not try to use CUDA VMM"                OFF)
+option(GGML_CUDA_FA_ALL_QUANTS              "ggml: compile all quants for FlashAttention"     OFF)
+option(GGML_CUDA_USE_GRAPHS                 "ggml: use CUDA graphs (llama.cpp only)"          OFF)
+
+option(GGML_CURL                            "ggml: use libcurl to download model from an URL" OFF)
+option(GGML_HIPBLAS                         "ggml: use hipBLAS"                               OFF)
+option(GGML_HIP_UMA                         "ggml: use HIP unified memory architecture"       OFF)
+option(GGML_VULKAN                          "ggml: use Vulkan"                                OFF)
+option(GGML_VULKAN_CHECK_RESULTS            "ggml: run Vulkan op checks"                      OFF)
+option(GGML_VULKAN_DEBUG                    "ggml: enable Vulkan debug output"                OFF)
+option(GGML_VULKAN_MEMORY_DEBUG             "ggml: enable Vulkan memory debug output"         OFF)
+option(GGML_VULKAN_VALIDATE                 "ggml: enable Vulkan validation"                  OFF)
+option(GGML_VULKAN_RUN_TESTS                "ggml: run Vulkan tests"                          OFF)
+option(GGML_KOMPUTE                         "ggml: use Kompute"                               OFF)
+option(GGML_METAL                           "ggml: use Metal"                                 ${GGML_METAL_DEFAULT})
+option(GGML_METAL_NDEBUG                    "ggml: disable Metal debugging"                   OFF)
+option(GGML_METAL_SHADER_DEBUG              "ggml: compile Metal with -fno-fast-math"         OFF)
+option(GGML_METAL_EMBED_LIBRARY             "ggml: embed Metal library"                       ${GGML_METAL})
+set   (GGML_METAL_MACOSX_VERSION_MIN "" CACHE STRING
+                                            "ggml: metal minimum macOS version")
+set   (GGML_METAL_STD "" CACHE STRING       "ggml: metal standard version (-std flag)")
+option(GGML_OPENMP                          "ggml: use OpenMP"                                ON)
+option(GGML_RPC                             "ggml: use RPC"                                   OFF)
+option(GGML_SYCL                            "ggml: use SYCL"                                  OFF)
+option(GGML_SYCL_F16                        "ggml: use 16 bit floats for sycl calculations"   OFF)
+set   (GGML_SYCL_TARGET "INTEL" CACHE STRING
+                                            "ggml: sycl target device")
+
+# extra artifacts
+option(GGML_BUILD_TESTS    "ggml: build tests"    ${GGML_STANDALONE})
+option(GGML_BUILD_EXAMPLES "ggml: build examples" ${GGML_STANDALONE})
+
+#
+# dependencies
+#
+
+set(CMAKE_C_STANDARD 11)
+set(CMAKE_C_STANDARD_REQUIRED true)
+
+if (GGML_SYCL)
+    set(CMAKE_CXX_STANDARD 17)
+else()
+    set(CMAKE_CXX_STANDARD 11)
+endif()
+set(CMAKE_CXX_STANDARD_REQUIRED true)
+
+set(THREADS_PREFER_PTHREAD_FLAG ON)
+
+find_package(Threads REQUIRED)
+
+#
+# build the library
+#
+
+add_subdirectory(src)
+
+#
+# tests and examples
+#
+
+if (GGML_BUILD_TESTS)
+    enable_testing()
+    add_subdirectory(tests)
+endif ()
+
+if (GGML_BUILD_EXAMPLES)
+    add_subdirectory(examples)
+endif ()
+
+#
+# install
+#
+
+include(GNUInstallDirs)
+include(CMakePackageConfigHelpers)
+
+# all public headers
+set(GGML_PUBLIC_HEADERS
+    include/ggml.h
+    include/ggml-alloc.h
+    include/ggml-backend.h
+    include/ggml-blas.h
+    include/ggml-cuda.h
+    include/ggml.h
+    include/ggml-kompute.h
+    include/ggml-metal.h
+    include/ggml-rpc.h
+    include/ggml-sycl.h
+    include/ggml-vulkan.h)
+
+set_target_properties(ggml PROPERTIES PUBLIC_HEADER "${GGML_PUBLIC_HEADERS}")
+#if (GGML_METAL)
+#    set_target_properties(ggml PROPERTIES RESOURCE "${CMAKE_CURRENT_SOURCE_DIR}/src/ggml-metal.metal")
+#endif()
+install(TARGETS ggml PUBLIC_HEADER)
+
+if (BUILD_SHARED_LIBS)
+    install(TARGETS ggml LIBRARY)
+endif()
+
+if (GGML_METAL)
+    install(
+        FILES src/ggml-metal.metal
+        PERMISSIONS
+            OWNER_READ
+            OWNER_WRITE
+            GROUP_READ
+            WORLD_READ
+        DESTINATION ${CMAKE_INSTALL_BINDIR})
+
+    if (NOT GGML_METAL_EMBED_LIBRARY)
+        install(
+            FILES ${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/default.metallib
+            DESTINATION ${CMAKE_INSTALL_BINDIR}
+        )
+    endif()
+endif()
+
+if (GGML_STANDALONE)
+    configure_file(${CMAKE_CURRENT_SOURCE_DIR}/ggml.pc.in
+        ${CMAKE_CURRENT_BINARY_DIR}/ggml.pc
+        @ONLY)
+
+    install(FILES ${CMAKE_CURRENT_BINARY_DIR}/ggml.pc
+        DESTINATION share/pkgconfig)
+endif()