Merge pull request #2 from ilya-lavrenov/jinja-integration-pavel

Integrate JinjaCpp

CMakeLists.txt

@@ -1,10 +1,10 @@
-# Copyright (C) 2018-2023 Intel Corporation
+# Copyright (C) 2018-2024 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 #
 
 cmake_minimum_required(VERSION 3.15)
-project(openvino_genai)
 
+project(openvino_genai)
 
 add_subdirectory(src)
 add_subdirectory(text_generation/causal_lm/cpp)

src/CMakeLists.txt

@@ -1,6 +1,6 @@
-# Copyright (C) 2018-2023 Intel Corporation
+# Copyright (C) 2018-2024 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 #
 
-add_subdirectory(python-bindings)
 add_subdirectory(cpp)
+add_subdirectory(python)

src/cpp/CMakeLists.txt

@@ -1,25 +1,60 @@
-# Generate Pipeline library
+# Copyright (C) 2018-2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+#
 
-set(JINJA2CPP_DEPS_MODE internal)
+# Dependencies
+
+include(FetchContent)
+
+FetchContent_Declare(nlohmann_json
+    URL https://github.com/nlohmann/json/archive/refs/tags/v3.11.3.tar.gz
+    URL_HASH SHA256=0d8ef5af7f9794e3263480193c491549b2ba6cc74bb018906202ada498a79406)
+FetchContent_MakeAvailable(nlohmann_json)
+
+function(ov_genai_build_jinja2cpp)
+    FetchContent_Declare(jinja2cpp
+        URL https://github.com/ilya-lavrenov/Jinja2Cpp/archive/a5d002cbf44469775556daea14ba3ccdba1e365a.tar.gz
+        URL_HASH SHA256=5aa5378d9acf3c44dfb607fd7f16f48b17ffa6495c219957901e9191ffe28900)
+
+    FetchContent_GetProperties(jinja2cpp)
+    if(NOT jinja2cpp_POPULATED)
+        FetchContent_Populate(jinja2cpp)
+
+        set(BUILD_SHARED_LIBS OFF)
+        set(JINJA2CPP_INSTALL OFF CACHE BOOL "")
+        set(JINJA2CPP_CXX_STANDARD 17 CACHE STRING "")
+        set(JINJA2CPP_BUILD_SHARED OFF CACHE BOOL "")
+        set(JINJA2CPP_USE_REGEX "std" CACHE STRING "")
+        set(JINJA2CPP_WITH_JSON_BINDINGS "none" CACHE STRING "")
+        set(JINJA2CPP_STRICT_WARNINGS OFF CACHE BOOL "")
+        set(JINJA2CPP_PIC ON CACHE BOOL "")
+
+        add_subdirectory("${jinja2cpp_SOURCE_DIR}" "${jinja2cpp_BINARY_DIR}" EXCLUDE_FROM_ALL)
+    endif()
+endfunction()
+
+ov_genai_build_jinja2cpp()
 
 add_subdirectory(../../thirdparty/openvino_tokenizers/ "${CMAKE_CURRENT_BINARY_DIR}/openvino_tokenizers/")
-add_subdirectory(../../thirdparty/nlohmann_json/ "${CMAKE_CURRENT_BINARY_DIR}/nlohmann_json/")
 
-# todo: remove hardcodes and make submodule work
-# include_directories($ENV{HOME}/opt/jinja2cpp/include)
-# add_subdirectory(../../../thirdparty/Jinja2Cpp/ "${CMAKE_CURRENT_BINARY_DIR}/Jinja2Cpp/")
-# include_directories(../../../thirdparty/inja/include/Jinja2Cpp)
+find_package(OpenVINO REQUIRED COMPONENTS Runtime)
+
+# Library
 
+file(GLOB SOURCE_FILES "${CMAKE_CURRENT_SOURCE_DIR}/src/*.cpp")
 
 set(TARGET_NAME generate_pipeline_lib)
-file(GLOB SOURCE_FILES "src/*.cpp")
 add_library(${TARGET_NAME} SHARED ${SOURCE_FILES})
-target_include_directories(${TARGET_NAME} PRIVATE ../../text_generation/causal_lm/cpp/)
-target_include_directories(${TARGET_NAME} PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/include)
-find_package(OpenVINO REQUIRED COMPONENTS Runtime)
-target_link_libraries(${TARGET_NAME} PUBLIC openvino::runtime)
-target_link_libraries(${TARGET_NAME} PUBLIC nlohmann_json::nlohmann_json)
+
+target_include_directories(${TARGET_NAME}
+    # TODO: remove it, because beam_search algo should not be exposed to end users
+    PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/../../text_generation/causal_lm/cpp/
+    PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/include)
+
+target_link_libraries(${TARGET_NAME} PUBLIC openvino::runtime PRIVATE nlohmann_json::nlohmann_json jinja2cpp)
+
 target_compile_definitions(${TARGET_NAME} PRIVATE OPENVINO_TOKENIZERS_PATH=\"$<TARGET_FILE:openvino_tokenizers>\")
-# target_link_libraries(${TARGET_NAME} PRIVATE $ENV{HOME}/opt/jinja2cpp/lib/static/libjinja2cpp.a)  # todo: remove hardcode
-set_target_properties(${TARGET_NAME} PROPERTIES CXX_STANDARD 17)
-set_target_properties(${TARGET_NAME} PROPERTIES CXX_STANDARD_REQUIRED ON)
+
+set_target_properties(${TARGET_NAME} PROPERTIES
+    CXX_STANDARD_REQUIRED ON
+    CXX_STANDARD 17)

src/cpp/src/llm_pipeline.cpp

@@ -11,9 +11,9 @@
 #include "utils.hpp"
 #include <nlohmann/json.hpp>
 
-// #include <jinja2cpp/template.h>
-// #include <jinja2cpp/template_env.h>
-// #include "generation_config.hpp"
+#include <jinja2cpp/template.h>
+#include <jinja2cpp/template_env.h>
+#include "openvino/genai/generation_config.hpp"
 
 
 namespace ov {
@@ -313,29 +313,21 @@ std::string ov::LLMPipeline::apply_chat_template(std::string prompt, std::string
 }
 
 std::string ov::LLMPipeline::LLMPipelineImpl::apply_chat_template(std::string prompt, std::string role) const {
-    // todo: temporary disable for easier and faster build
-    // jinja2::TemplateEnv env;
-    // env.GetSettings().lstripBlocks = true;
-    // env.GetSettings().trimBlocks = true;
-    // jinja2::Template tpl(&env);
-    // tpl.Load(m_chat_template);
+    jinja2::TemplateEnv env;
+    env.GetSettings().lstripBlocks = true;
+    env.GetSettings().trimBlocks = true;
+    jinja2::Template tpl(&env);
+    tpl.Load(m_chat_template);
 
-    // jinja2::ValuesMap message {{"role", role}, {"content", prompt}};
-    // jinja2::ValuesMap params = {
-    //     {"messages", jinja2::ValuesList({message})},
-    //     {"bos_token",  "<s>"},
-    //     {"eos_token", "</s>"},  // todo: load from config
-    //     {"add_generation_prompt", true},
-    // };
+    jinja2::ValuesMap message {{"role", role}, {"content", prompt}};
+    jinja2::ValuesMap params = {
+        {"messages", jinja2::ValuesList({message})},
+        {"bos_token",  "<s>"},
+        {"eos_token", "</s>"},  // todo: load from config
+        {"add_generation_prompt", true},
+    };
 
-    // return tpl.RenderAsString(params).value();
-
-    std::stringstream result_prompt;
-    result_prompt << "<|user|>\n" << prompt << "</s>\n<|assistant|>\n";  // hardcode template for TinyLlama
-    // result_prompt << "<bos><start_of_turn>user\n" << prompt << "<end_of_turn>\n<start_of_turn>model";  // Gemma-7b-it
-    // result_prompt << "<s>[INST] " << input << " [/INST]";  // LLama-2-7b
-
-    return result_prompt.str();
+    return tpl.RenderAsString(params).value();
 }
 
 void ov::LLMPipeline::start_chat() {

src/python-bindings/CMakeLists.txt → src/python/CMakeLists.txt

@@ -1,3 +1,7 @@
+# Copyright (C) 2018-2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+#
+
 include(FetchContent)
 FetchContent_Declare(
     pybind11