cui-llama.rn 1.5.0 → 1.6.1

package/cpp/unicode.h

@@ -1,66 +1,66 @@
-#pragma once
-
-#include <cstdint>
-#include <string>
-#include <vector>
-
-struct unicode_cpt_flags {
-    enum {
-        UNDEFINED       = 0x0001,
-        NUMBER          = 0x0002,  // regex: \p{N}
-        LETTER          = 0x0004,  // regex: \p{L}
-        SEPARATOR       = 0x0008,  // regex: \p{Z}
-        ACCENT_MARK     = 0x0010,  // regex: \p{M}
-        PUNCTUATION     = 0x0020,  // regex: \p{P}
-        SYMBOL          = 0x0040,  // regex: \p{S}
-        CONTROL         = 0x0080,  // regex: \p{C}
-        MASK_CATEGORIES = 0x00FF,
-    };
-
-    // codepoint type
-    uint16_t is_undefined   : 1;
-    uint16_t is_number      : 1;  // regex: \p{N}
-    uint16_t is_letter      : 1;  // regex: \p{L}
-    uint16_t is_separator   : 1;  // regex: \p{Z}
-    uint16_t is_accent_mark : 1;  // regex: \p{M}
-    uint16_t is_punctuation : 1;  // regex: \p{P}
-    uint16_t is_symbol      : 1;  // regex: \p{S}
-    uint16_t is_control     : 1;  // regex: \p{C}
-    // helper flags
-    uint16_t is_whitespace  : 1;  // regex: \s
-    uint16_t is_lowercase   : 1;
-    uint16_t is_uppercase   : 1;
-    uint16_t is_nfd         : 1;
-
-    // decode from uint16
-    inline unicode_cpt_flags(const uint16_t flags = 0) {
-        *reinterpret_cast<uint16_t*>(this) = flags;
-    }
-
-    inline uint16_t as_uint() const {
-        return *reinterpret_cast<const uint16_t*>(this);
-    }
-
-    inline uint16_t category_flag() const {
-        return this->as_uint() & MASK_CATEGORIES;
-    }
-};
-
-size_t unicode_len_utf8(char src);
-
-std::string unicode_cpt_to_utf8  (uint32_t cpt);
-uint32_t    unicode_cpt_from_utf8(const std::string & utf8, size_t & offset);
-
-std::vector<uint32_t> unicode_cpts_from_utf8(const std::string & utf8);
-
-std::vector<uint32_t> unicode_cpts_normalize_nfd(const std::vector<uint32_t> & cpts);
-
-unicode_cpt_flags unicode_cpt_flags_from_cpt (uint32_t cpt);
-unicode_cpt_flags unicode_cpt_flags_from_utf8(const std::string & utf8);
-
-std::string unicode_byte_to_utf8(uint8_t byte);
-uint8_t     unicode_utf8_to_byte(const std::string & utf8);
-
-uint32_t unicode_tolower(uint32_t cpt);
-
-std::vector<std::string> unicode_regex_split(const std::string & text, const std::vector<std::string> & regex_exprs);
+#pragma once
+
+#include <cstdint>
+#include <string>
+#include <vector>
+
+struct unicode_cpt_flags {
+    enum {
+        UNDEFINED       = 0x0001,
+        NUMBER          = 0x0002,  // regex: \p{N}
+        LETTER          = 0x0004,  // regex: \p{L}
+        SEPARATOR       = 0x0008,  // regex: \p{Z}
+        ACCENT_MARK     = 0x0010,  // regex: \p{M}
+        PUNCTUATION     = 0x0020,  // regex: \p{P}
+        SYMBOL          = 0x0040,  // regex: \p{S}
+        CONTROL         = 0x0080,  // regex: \p{C}
+        MASK_CATEGORIES = 0x00FF,
+    };
+
+    // codepoint type
+    uint16_t is_undefined   : 1;
+    uint16_t is_number      : 1;  // regex: \p{N}
+    uint16_t is_letter      : 1;  // regex: \p{L}
+    uint16_t is_separator   : 1;  // regex: \p{Z}
+    uint16_t is_accent_mark : 1;  // regex: \p{M}
+    uint16_t is_punctuation : 1;  // regex: \p{P}
+    uint16_t is_symbol      : 1;  // regex: \p{S}
+    uint16_t is_control     : 1;  // regex: \p{C}
+    // helper flags
+    uint16_t is_whitespace  : 1;  // regex: \s
+    uint16_t is_lowercase   : 1;
+    uint16_t is_uppercase   : 1;
+    uint16_t is_nfd         : 1;
+
+    // decode from uint16
+    inline unicode_cpt_flags(const uint16_t flags = 0) {
+        *reinterpret_cast<uint16_t*>(this) = flags;
+    }
+
+    inline uint16_t as_uint() const {
+        return *reinterpret_cast<const uint16_t*>(this);
+    }
+
+    inline uint16_t category_flag() const {
+        return this->as_uint() & MASK_CATEGORIES;
+    }
+};
+
+size_t unicode_len_utf8(char src);
+
+std::string unicode_cpt_to_utf8  (uint32_t cpt);
+uint32_t    unicode_cpt_from_utf8(const std::string & utf8, size_t & offset);
+
+std::vector<uint32_t> unicode_cpts_from_utf8(const std::string & utf8);
+
+std::vector<uint32_t> unicode_cpts_normalize_nfd(const std::vector<uint32_t> & cpts);
+
+unicode_cpt_flags unicode_cpt_flags_from_cpt (uint32_t cpt);
+unicode_cpt_flags unicode_cpt_flags_from_utf8(const std::string & utf8);
+
+std::string unicode_byte_to_utf8(uint8_t byte);
+uint8_t     unicode_utf8_to_byte(const std::string & utf8);
+
+uint32_t unicode_tolower(uint32_t cpt);
+
+std::vector<std::string> unicode_regex_split(const std::string & text, const std::vector<std::string> & regex_exprs);

package/ios/CMakeLists.txt

@@ -1,108 +1,119 @@
-cmake_minimum_required(VERSION 3.16)
-project(rnllama VERSION 1.0.0 LANGUAGES CXX C)
-
-set(CMAKE_CXX_STANDARD 17)
-set(CMAKE_CXX_STANDARD_REQUIRED ON)
-
-# iOS specific settings
-set(CMAKE_OSX_DEPLOYMENT_TARGET 13.0)
-set(CMAKE_XCODE_ATTRIBUTE_ENABLE_BITCODE NO)
-
-# Dependencies and compile options
-add_definitions(
-    -DNDEBUG
-    -DO3
-    -DLM_GGML_USE_CPU
-    -DLM_GGML_USE_ACCELERATE
-    -DLM_GGML_USE_METAL
-    -DLM_GGML_METAL_USE_BF16
-)
-
-set(SOURCE_DIR ${CMAKE_CURRENT_SOURCE_DIR}/../cpp)
-
-# Define public headers
-set(PUBLIC_HEADERS
-    ${SOURCE_DIR}/rn-llama.h
-    ${SOURCE_DIR}/llama.h
-    ${SOURCE_DIR}/llama-impl.h
-    ${SOURCE_DIR}/ggml.h
-)
-
-# Create library target
-add_library(rnllama SHARED
-    ${SOURCE_DIR}/ggml.c
-    ${SOURCE_DIR}/ggml-alloc.c
-    ${SOURCE_DIR}/ggml-backend.cpp
-    ${SOURCE_DIR}/ggml-backend-reg.cpp
-    ${SOURCE_DIR}/ggml-cpu.c
-    ${SOURCE_DIR}/ggml-cpu.cpp
-    ${SOURCE_DIR}/ggml-cpu-aarch64.cpp
-    ${SOURCE_DIR}/ggml-cpu-quants.c
-    ${SOURCE_DIR}/ggml-cpu-traits.cpp
-    ${SOURCE_DIR}/ggml-metal.m
-    ${SOURCE_DIR}/ggml-opt.cpp
-    ${SOURCE_DIR}/ggml-threading.cpp
-    ${SOURCE_DIR}/ggml-quants.c
-    ${SOURCE_DIR}/gguf.cpp
-    ${SOURCE_DIR}/log.cpp
-    ${SOURCE_DIR}/llama-impl.cpp
-    ${SOURCE_DIR}/llama-grammar.cpp
-    ${SOURCE_DIR}/llama-sampling.cpp
-    ${SOURCE_DIR}/llama-vocab.cpp
-    ${SOURCE_DIR}/llama-adapter.cpp
-    ${SOURCE_DIR}/llama-chat.cpp
-    ${SOURCE_DIR}/llama-context.cpp
-    ${SOURCE_DIR}/llama-kv-cache.cpp
-    ${SOURCE_DIR}/llama-arch.cpp
-    ${SOURCE_DIR}/llama-batch.cpp
-    ${SOURCE_DIR}/llama-cparams.cpp
-    ${SOURCE_DIR}/llama-hparams.cpp
-    ${SOURCE_DIR}/llama.cpp
-    ${SOURCE_DIR}/llama-model.cpp
-    ${SOURCE_DIR}/llama-model-loader.cpp
-    ${SOURCE_DIR}/llama-mmap.cpp
-    ${SOURCE_DIR}/llama-vocab.cpp
-    ${SOURCE_DIR}/llama-memory.cpp
-    ${SOURCE_DIR}/llama-io.cpp
-    ${SOURCE_DIR}/llama-graph.cpp
-    ${SOURCE_DIR}/sampling.cpp
-    ${SOURCE_DIR}/unicode-data.cpp
-    ${SOURCE_DIR}/unicode.cpp
-    ${SOURCE_DIR}/sgemm.cpp
-    ${SOURCE_DIR}/common.cpp
-    ${SOURCE_DIR}/chat.cpp
-    ${SOURCE_DIR}/json-schema-to-grammar.cpp
-    ${SOURCE_DIR}/minja/minja.hpp
-    ${SOURCE_DIR}/minja/chat-template.hpp
-    ${SOURCE_DIR}/json.hpp
-    ${SOURCE_DIR}/amx/amx.cpp
-    ${SOURCE_DIR}/amx/mmq.cpp
-    ${SOURCE_DIR}/rn-llama.cpp
-)
-
-# Setup include directories
-target_include_directories(rnllama
-    PUBLIC
-        $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/../cpp>
-        $<INSTALL_INTERFACE:include>
-)
-
-# Link required frameworks
-target_link_libraries(rnllama PRIVATE
-    "-framework Accelerate"
-    "-framework Foundation"
-    "-framework Metal"
-    "-framework MetalKit"
-)
-
-# Set properties for framework
-set_target_properties(rnllama PROPERTIES
-    MACOSX_FRAMEWORK_IDENTIFIER "com.rnllama"
-    MACOSX_FRAMEWORK_BUNDLE_VERSION 1.0.0
-    MACOSX_FRAMEWORK_SHORT_VERSION_STRING 1.0.0
-    FRAMEWORK TRUE
-    FRAMEWORK_VERSION 1.0.0
-    VERSION 1.0.0
-    PUBLIC_HEADER "${PUBLIC_HEADERS}"
-    XCODE_ATTRIBUTE_CLANG_ENABLE_OBJC_ARC NO
-)
+cmake_minimum_required(VERSION 3.16)
+project(rnllama VERSION 1.0.0 LANGUAGES CXX C)
+
+find_program(CCACHE_FOUND ccache)
+if(CCACHE_FOUND)
+  set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE ccache)
+  set_property(GLOBAL PROPERTY RULE_LAUNCH_LINK ccache)
+endif(CCACHE_FOUND)
+
+set(CMAKE_CXX_STANDARD 17)
+set(CMAKE_CXX_STANDARD_REQUIRED ON)
+
+# iOS specific settings
+set(CMAKE_OSX_DEPLOYMENT_TARGET 13.0)
+set(CMAKE_XCODE_ATTRIBUTE_ENABLE_BITCODE NO)
+
+# Dependencies and compile options
+add_definitions(
+    -DNDEBUG
+    -DO3
+    -DLM_GGML_USE_CPU
+    -DLM_GGML_USE_ACCELERATE
+    -DLM_GGML_USE_METAL
+    -DLM_GGML_METAL_USE_BF16
+)
+
+set(SOURCE_DIR ${CMAKE_CURRENT_SOURCE_DIR}/../cpp)
+
+# Define public headers
+set(PUBLIC_HEADERS
+    ${SOURCE_DIR}/rn-llama.h
+    ${SOURCE_DIR}/llama.h
+    ${SOURCE_DIR}/llama-impl.h
+    ${SOURCE_DIR}/ggml.h
+)
+
+# Create library target
+add_library(rnllama SHARED
+    ${SOURCE_DIR}/ggml.c
+    ${SOURCE_DIR}/ggml-alloc.c
+    ${SOURCE_DIR}/ggml-backend.cpp
+    ${SOURCE_DIR}/ggml-backend-reg.cpp
+    ${SOURCE_DIR}/ggml-cpu/amx/amx.cpp
+    ${SOURCE_DIR}/ggml-cpu/amx/mmq.cpp
+    ${SOURCE_DIR}/ggml-cpu/ggml-cpu.c
+    ${SOURCE_DIR}/ggml-cpu/ggml-cpu.cpp
+    ${SOURCE_DIR}/ggml-cpu/ggml-cpu-aarch64.cpp
+    ${SOURCE_DIR}/ggml-cpu/ggml-cpu-quants.c
+    ${SOURCE_DIR}/ggml-cpu/ggml-cpu-traits.cpp
+    ${SOURCE_DIR}/ggml-cpu/unary-ops.cpp
+    ${SOURCE_DIR}/ggml-cpu/binary-ops.cpp
+    ${SOURCE_DIR}/ggml-cpu/sgemm.cpp
+    ${SOURCE_DIR}/ggml-cpu/vec.cpp
+    ${SOURCE_DIR}/ggml-cpu/ops.cpp
+    ${SOURCE_DIR}/ggml-metal.m
+    ${SOURCE_DIR}/ggml-opt.cpp
+    ${SOURCE_DIR}/ggml-threading.cpp
+    ${SOURCE_DIR}/ggml-quants.c
+    ${SOURCE_DIR}/gguf.cpp
+    ${SOURCE_DIR}/log.cpp
+    ${SOURCE_DIR}/llama-impl.cpp
+    ${SOURCE_DIR}/llama-grammar.cpp
+    ${SOURCE_DIR}/llama-sampling.cpp
+    ${SOURCE_DIR}/llama-vocab.cpp
+    ${SOURCE_DIR}/llama-adapter.cpp
+    ${SOURCE_DIR}/llama-chat.cpp
+    ${SOURCE_DIR}/llama-context.cpp
+    ${SOURCE_DIR}/llama-kv-cache.cpp
+    ${SOURCE_DIR}/llama-arch.cpp
+    ${SOURCE_DIR}/llama-batch.cpp
+    ${SOURCE_DIR}/llama-cparams.cpp
+    ${SOURCE_DIR}/llama-hparams.cpp
+    ${SOURCE_DIR}/llama.cpp
+    ${SOURCE_DIR}/llama-model.cpp
+    ${SOURCE_DIR}/llama-model-loader.cpp
+    ${SOURCE_DIR}/llama-mmap.cpp
+    ${SOURCE_DIR}/llama-vocab.cpp
+    ${SOURCE_DIR}/llama-memory.cpp
+    ${SOURCE_DIR}/llama-io.cpp
+    ${SOURCE_DIR}/llama-graph.cpp
+    ${SOURCE_DIR}/sampling.cpp
+    ${SOURCE_DIR}/unicode-data.cpp
+    ${SOURCE_DIR}/unicode.cpp
+    ${SOURCE_DIR}/common.cpp
+    ${SOURCE_DIR}/chat.cpp
+    ${SOURCE_DIR}/json-schema-to-grammar.cpp
+    ${SOURCE_DIR}/minja/minja.hpp
+    ${SOURCE_DIR}/minja/chat-template.hpp
+    ${SOURCE_DIR}/json.hpp
+    ${SOURCE_DIR}/rn-llama.cpp
+)
+
+# Setup include directories
+target_include_directories(rnllama
+    PUBLIC
+        $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/../cpp>
+        $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/../cpp/ggml-cpu>
+        $<INSTALL_INTERFACE:include>
+)
+
+# Link required frameworks
+target_link_libraries(rnllama PRIVATE
+    "-framework Accelerate"
+    "-framework Foundation"
+    "-framework Metal"
+    "-framework MetalKit"
+)
+
+# Set properties for framework
+set_target_properties(rnllama PROPERTIES
+    MACOSX_FRAMEWORK_IDENTIFIER "com.rnllama"
+    MACOSX_FRAMEWORK_BUNDLE_VERSION 1.0.0
+    MACOSX_FRAMEWORK_SHORT_VERSION_STRING 1.0.0
+    FRAMEWORK TRUE
+    FRAMEWORK_VERSION 1.0.0
+    VERSION 1.0.0
+    PUBLIC_HEADER "${PUBLIC_HEADERS}"
+    XCODE_ATTRIBUTE_CLANG_ENABLE_OBJC_ARC NO
+)

package/ios/RNLlama.h

@@ -1,7 +1,13 @@
-#import <React/RCTEventEmitter.h>
-#import <React/RCTBridgeModule.h>
-
-// TODO: Use RNLlamaSpec (Need to refactor NSDictionary usage)
-@interface RNLlama : RCTEventEmitter <RCTBridgeModule>
-
-@end
+#import <React/RCTEventEmitter.h>
+#import <React/RCTBridgeModule.h>
+
+#if RNLLAMA_BUILD_FROM_SOURCE
+#import "json.hpp"
+#else
+#import <rnllama/json.hpp>
+#endif
+
+// TODO: Use RNLlamaSpec (Need to refactor NSDictionary usage)
+@interface RNLlama : RCTEventEmitter <RCTBridgeModule>
+
+@end