Merge bb0b2c4f56 into 32d6ee6385

ggml/src/CMakeLists.txt

@@ -234,7 +234,6 @@ function(ggml_add_backend_library backend)
         # write the shared library to the output directory
         set_target_properties(${backend} PROPERTIES LIBRARY_OUTPUT_DIRECTORY ${CMAKE_RUNTIME_OUTPUT_DIRECTORY})
         target_compile_definitions(${backend} PRIVATE GGML_BACKEND_DL)
-        add_dependencies(ggml ${backend})
     else()
         add_library(${backend} ${ARGN})
         target_link_libraries(ggml PUBLIC ${backend})

ggml/src/ggml-backend-reg.cpp

@@ -66,26 +66,6 @@
 #include "ggml-kompute.h"
 #endif
 
-// disable C++17 deprecation warning for std::codecvt_utf8
-#if defined(__clang__)
-#    pragma clang diagnostic push
-#    pragma clang diagnostic ignored "-Wdeprecated-declarations"
-#endif
-
-static std::wstring utf8_to_utf16(const std::string & str) {
-    std::wstring_convert<std::codecvt_utf8_utf16<wchar_t>> converter;
-    return converter.from_bytes(str);
-}
-
-static std::string utf16_to_utf8(const std::wstring & str) {
-    std::wstring_convert<std::codecvt_utf8_utf16<wchar_t>> converter;
-    return converter.to_bytes(str);
-}
-
-#if defined(__clang__)
-#    pragma clang diagnostic pop
-#endif
-
 #ifdef _WIN32
 
 using dl_handle = std::remove_pointer_t<HMODULE>;
@@ -108,6 +88,11 @@ static dl_handle * dl_load_library(const std::wstring & path) {
     return handle;
 }
 
+static dl_handle * dl_load_library(const std::string & path) {
+    std::wstring_convert<std::codecvt_utf8_utf16<wchar_t>> converter;
+    return dl_load_library(converter.from_bytes(path));
+}
+
 static void * dl_get_sym(dl_handle * handle, const char * name) {
     DWORD old_mode = SetErrorMode(SEM_FAILCRITICALERRORS);
     SetErrorMode(old_mode | SEM_FAILCRITICALERRORS);
@@ -129,8 +114,8 @@ struct dl_handle_deleter {
     }
 };
 
-static void * dl_load_library(const std::wstring & path) {
-    dl_handle * handle = dlopen(utf16_to_utf8(path).c_str(), RTLD_NOW | RTLD_LOCAL);
+static void * dl_load_library(const std::string & path) {
+    dl_handle * handle = dlopen(path.c_str(), RTLD_NOW | RTLD_LOCAL);
 
     return handle;
 }
@@ -217,11 +202,11 @@ struct ggml_backend_registry {
         devices.push_back(device);
     }
 
-    ggml_backend_reg_t load_backend(const std::wstring & path, bool silent) {
+    ggml_backend_reg_t load_backend(const char * path, bool silent) {
         dl_handle_ptr handle { dl_load_library(path) };
         if (!handle) {
             if (!silent) {
-                GGML_LOG_ERROR("%s: failed to load %s\n", __func__, utf16_to_utf8(path).c_str());
+                GGML_LOG_ERROR("%s: failed to load %s\n", __func__, path);
             }
             return nullptr;
         }
@@ -229,7 +214,7 @@ struct ggml_backend_registry {
         auto score_fn = (ggml_backend_score_t) dl_get_sym(handle.get(), "ggml_backend_score");
         if (score_fn && score_fn() == 0) {
             if (!silent) {
-                GGML_LOG_INFO("%s: backend %s is not supported on this system\n", __func__, utf16_to_utf8(path).c_str());
+                GGML_LOG_INFO("%s: backend %s is not supported on this system\n", __func__, path);
             }
             return nullptr;
         }
@@ -237,7 +222,7 @@ struct ggml_backend_registry {
         auto backend_init_fn = (ggml_backend_init_t) dl_get_sym(handle.get(), "ggml_backend_init");
         if (!backend_init_fn) {
             if (!silent) {
-                GGML_LOG_ERROR("%s: failed to find ggml_backend_init in %s\n", __func__, utf16_to_utf8(path).c_str());
+                GGML_LOG_ERROR("%s: failed to find ggml_backend_init in %s\n", __func__, path);
             }
             return nullptr;
         }
@@ -246,16 +231,16 @@ struct ggml_backend_registry {
         if (!reg || reg->api_version != GGML_BACKEND_API_VERSION) {
             if (!silent) {
                 if (!reg) {
-                    GGML_LOG_ERROR("%s: failed to initialize backend from %s: ggml_backend_init returned NULL\n", __func__, utf16_to_utf8(path).c_str());
+                    GGML_LOG_ERROR("%s: failed to initialize backend from %s: ggml_backend_init returned NULL\n", __func__, path);
                 } else {
                     GGML_LOG_ERROR("%s: failed to initialize backend from %s: incompatible API version (backend: %d, current: %d)\n",
-                        __func__, utf16_to_utf8(path).c_str(), reg->api_version, GGML_BACKEND_API_VERSION);
+                        __func__, path, reg->api_version, GGML_BACKEND_API_VERSION);
                 }
             }
             return nullptr;
         }
 
-        GGML_LOG_INFO("%s: loaded %s backend from %s\n", __func__, ggml_backend_reg_name(reg), utf16_to_utf8(path).c_str());
+        GGML_LOG_INFO("%s: loaded %s backend from %s\n", __func__, ggml_backend_reg_name(reg), path);
 
         register_backend(reg, std::move(handle));
 
@@ -391,14 +376,14 @@ ggml_backend_t ggml_backend_init_best(void) {
 
 // Dynamic loading
 ggml_backend_reg_t ggml_backend_load(const char * path) {
-    return get_reg().load_backend(utf8_to_utf16(path), false);
+    return get_reg().load_backend(path, false);
 }
 
 void ggml_backend_unload(ggml_backend_reg_t reg) {
     get_reg().unload_backend(reg, true);
 }
 
-static std::wstring get_executable_path() {
+static std::string get_executable_path() {
 #if defined(__APPLE__)
     // get executable path
     std::vector<char> path;
@@ -416,7 +401,7 @@ static std::wstring get_executable_path() {
     if (last_slash != std::string::npos) {
         base_path = base_path.substr(0, last_slash);
     }
-    return utf8_to_utf16(base_path + "/");
+    return base_path + "/";
 #elif defined(__linux__) || defined(__FreeBSD__)
     std::string base_path = ".";
     std::vector<char> path(1024);
@@ -442,63 +427,57 @@ static std::wstring get_executable_path() {
         path.resize(path.size() * 2);
     }
 
-    return utf8_to_utf16(base_path + "/");
+    return base_path + "/";
 #elif defined(_WIN32)
-    std::vector<wchar_t> path(MAX_PATH);
-    DWORD len = GetModuleFileNameW(NULL, path.data(), path.size());
+    std::vector<char> path(MAX_PATH);
+    DWORD len = GetModuleFileNameA(NULL, path.data(), path.size());
     if (len == 0) {
-        return {};
+        return "";
     }
-    std::wstring base_path(path.data(), len);
+    std::string base_path(path.data(), len);
     // remove executable name
     auto last_slash = base_path.find_last_of('\\');
     if (last_slash != std::string::npos) {
         base_path = base_path.substr(0, last_slash);
     }
-    return base_path + L"\\";
-#else
-    return {};
+    return base_path + "\\";
 #endif
 }
 
-static std::wstring backend_filename_prefix() {
+static std::string backend_filename_prefix() {
 #ifdef _WIN32
-    return L"ggml-";
+    return "ggml-";
 #else
-    return L"libggml-";
+    return "libggml-";
 #endif
 }
 
-static std::wstring backend_filename_suffix() {
+static std::string backend_filename_suffix() {
 #ifdef _WIN32
-    return L".dll";
+    return ".dll";
 #else
-    return L".so";
-#endif
-}
-
-static std::wstring path_separator() {
-#ifdef _WIN32
-    return L"\\";
-#else
-    return L"/";
+    return ".so";
 #endif
 }
 
 static ggml_backend_reg_t ggml_backend_load_best(const char * name, bool silent, const char * user_search_path) {
     // enumerate all the files that match [lib]ggml-name-*.[so|dll] in the search paths
      // TODO: search system paths
-    std::wstring file_prefix = backend_filename_prefix() + utf8_to_utf16(name) + L"-";
-    std::vector<std::wstring> search_paths;
+    std::string file_prefix = backend_filename_prefix() + name + "-";
+    std::vector<std::string> search_paths;
     if (user_search_path == nullptr) {
-        search_paths.push_back(L"." + path_separator());
+        search_paths.push_back("./");
         search_paths.push_back(get_executable_path());
     } else {
-        search_paths.push_back(utf8_to_utf16(user_search_path) + path_separator());
+#if defined(_WIN32)
+        search_paths.push_back(std::string(user_search_path) + "\\");
+#else
+        search_paths.push_back(std::string(user_search_path) + "/");
+#endif
     }
 
     int best_score = 0;
-    std::wstring best_path;
+    std::string best_path;
 
     namespace fs = std::filesystem;
     for (const auto & search_path : search_paths) {
@@ -508,27 +487,27 @@ static ggml_backend_reg_t ggml_backend_load_best(const char * name, bool silent,
         fs::directory_iterator dir_it(search_path, fs::directory_options::skip_permission_denied);
         for (const auto & entry : dir_it) {
             if (entry.is_regular_file()) {
-                std::wstring filename = entry.path().filename().wstring();
-                std::wstring ext = entry.path().extension().wstring();
+                std::string filename = entry.path().filename().string();
+                std::string ext = entry.path().extension().string();
                 if (filename.find(file_prefix) == 0 && ext == backend_filename_suffix()) {
-                    dl_handle_ptr handle { dl_load_library(entry.path().wstring()) };
+                    dl_handle_ptr handle { dl_load_library(entry.path().c_str()) };
                     if (!handle && !silent) {
-                        GGML_LOG_ERROR("%s: failed to load %s\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str());
+                        GGML_LOG_ERROR("%s: failed to load %s\n", __func__, entry.path().string().c_str());
                     }
                     if (handle) {
                         auto score_fn = (ggml_backend_score_t) dl_get_sym(handle.get(), "ggml_backend_score");
                         if (score_fn) {
                             int s = score_fn();
 #ifndef NDEBUG
-                            GGML_LOG_DEBUG("%s: %s score: %d\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str(), s);
+                            GGML_LOG_DEBUG("%s: %s score: %d\n", __func__, entry.path().string().c_str(), s);
 #endif
                             if (s > best_score) {
                                 best_score = s;
-                                best_path = entry.path().wstring();
+                                best_path = entry.path().string();
                             }
                         } else {
                             if (!silent) {
-                                GGML_LOG_INFO("%s: failed to find ggml_backend_score in %s\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str());
+                                GGML_LOG_INFO("%s: failed to find ggml_backend_score in %s\n", __func__, entry.path().string().c_str());
                             }
                         }
                     }
@@ -540,15 +519,15 @@ static ggml_backend_reg_t ggml_backend_load_best(const char * name, bool silent,
     if (best_score == 0) {
         // try to load the base backend
         for (const auto & search_path : search_paths) {
-            std::wstring path = search_path + backend_filename_prefix() + utf8_to_utf16(name) + backend_filename_suffix();
+            std::string path = search_path + backend_filename_prefix() + name + backend_filename_suffix();
             if (fs::exists(path)) {
-                return get_reg().load_backend(path, silent);
+                return get_reg().load_backend(path.c_str(), silent);
             }
         }
         return nullptr;
     }
 
-    return get_reg().load_backend(best_path, silent);
+    return get_reg().load_backend(best_path.c_str(), silent);
 }
 
 void ggml_backend_load_all() {

ggml/src/ggml-cpu/CMakeLists.txt

@@ -135,20 +135,14 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
             endif()
 
             # show enabled features
-            if (CMAKE_HOST_SYSTEM_NAME STREQUAL "Windows")
-                set(FEAT_INPUT_FILE "NUL")
-            else()
-                set(FEAT_INPUT_FILE "/dev/null")
-            endif()
-
             execute_process(
                 COMMAND ${CMAKE_C_COMPILER} ${ARCH_FLAGS} -dM -E -
-                INPUT_FILE ${FEAT_INPUT_FILE}
+                INPUT_FILE "/dev/null"
                 OUTPUT_VARIABLE ARM_FEATURE
                 RESULT_VARIABLE ARM_FEATURE_RESULT
             )
             if (ARM_FEATURE_RESULT)
-                message(WARNING "Failed to get ARM features")
+                message(FATAL_ERROR "Failed to get ARM features")
             else()
                 foreach(feature DOTPROD SVE MATMUL_INT8 FMA FP16_VECTOR_ARITHMETIC)
                     string(FIND "${ARM_FEATURE}" "__ARM_FEATURE_${feature} 1" feature_pos)
@@ -323,11 +317,6 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
     target_compile_definitions(${GGML_CPU_NAME} PRIVATE ${ARCH_DEFINITIONS})
 
     if (GGML_BACKEND_DL)
-        if (GGML_NATIVE)
-            # the feature check relies on ARCH_DEFINITIONS, but it is not set with GGML_NATIVE
-            message(FATAL_ERROR "GGML_NATIVE is not compatible with GGML_BACKEND_DL, consider using GGML_CPU_ALL_VARIANTS")
-        endif()
-
         # The feature detection code is compiled as a separate target so that
         # it can be built without the architecture flags
         # Since multiple variants of the CPU backend may be included in the same

src/llama-vocab.cpp

@@ -1643,7 +1643,7 @@ bool llama_token_is_control_impl(const struct llama_vocab & vocab, llama_token t
 }
 
 llama_token llama_token_bos_impl(const struct llama_vocab & vocab) {
-    return vocab.type != LLAMA_VOCAB_TYPE_WPM ? vocab.special_bos_id : vocab.special_cls_id;
+    return vocab.special_bos_id;
 }
 
 llama_token llama_token_eos_impl(const struct llama_vocab & vocab) {

src/llama-vocab.h

@@ -57,7 +57,7 @@ struct llama_vocab {
     id special_unk_id  = 0;
     id special_sep_id  = LLAMA_TOKEN_NULL;
     id special_pad_id  = LLAMA_TOKEN_NULL;
-    id special_cls_id  = LLAMA_TOKEN_NULL; // TODO: revisit if this is really needed https://github.com/ggerganov/llama.cpp/pull/10930
+    id special_cls_id  = LLAMA_TOKEN_NULL;
     id special_mask_id = LLAMA_TOKEN_NULL;
 
     id linefeed_id = 13;

src/llama.cpp

@@ -1531,11 +1531,9 @@ static void llm_load_hparams(
                         switch (hparams.n_embd) {
                             case 384: model.type = e_model::MODEL_33M; break; // MiniLM-L12, bge-small
                             case 768: model.type = e_model::MODEL_109M; break; // bge-base
-                            default: model.type = e_model::MODEL_UNKNOWN;
                         } break;
                     case 24:
                         model.type = e_model::MODEL_335M; break; // bge-large
-                    default: model.type = e_model::MODEL_UNKNOWN;
                 }
             } break;
         case LLM_ARCH_JINA_BERT_V2:
@@ -1549,7 +1547,6 @@ static void llm_load_hparams(
                 switch (hparams.n_layer) {
                     case 4:  model.type = e_model::MODEL_33M;  break; // jina-embeddings-small
                     case 12: model.type = e_model::MODEL_137M; break; // jina-embeddings-base
-                    default: model.type = e_model::MODEL_UNKNOWN;
                 }
             } break;
         case LLM_ARCH_NOMIC_BERT:
@@ -1573,9 +1570,7 @@ static void llm_load_hparams(
                         switch (hparams.n_embd) {
                             case 2560: model.type = e_model::MODEL_3B; break;
                             case 4096: model.type = e_model::MODEL_7B; break;
-                            default: model.type = e_model::MODEL_UNKNOWN;
                         } break;
-                    default: model.type = e_model::MODEL_UNKNOWN;
                 }
 
                 // TODO: become GGUF KV parameter
@@ -2104,7 +2099,7 @@ static void llm_load_hparams(
                 ml.get_key(LLM_KV_ATTENTION_GROUPNORM_GROUPS, hparams.n_norm_groups);
                 ml.get_key(LLM_KV_ATTENTION_CAUSAL,           hparams.causal_attn);
             } break;
-        default: throw std::runtime_error("unsupported model architecture");
+        default: (void)0;
     }
 
     model.ftype = ml.ftype;