[CB] Align the speculative decoding sample with other ones (#1007)

Spin off of #907. Merge after head PR --------- Co-authored-by: Ilya Lavrenov <[email protected]>
openvinotoolkit · Oct 18, 2024 · b925196 · b925196
1 parent 2240669
commit b925196
Show file tree

Hide file tree

Showing 15 changed files with 133 additions and 191 deletions.
diff --git a/.github/workflows/causal_lm_cpp.yml b/.github/workflows/causal_lm_cpp.yml
@@ -422,8 +422,16 @@ jobs:
       - name: run and compare
         run: |
           source ./ov/setupvars.sh
-          ./build/samples/cpp/speculative_decoding_lm/speculative_decoding_lm -a ./dolly-v2-3b/ -m ./dolly-v2-7b/ -n 5
-          ./build/samples/cpp/speculative_decoding_lm/continuous_batching_speculative_decoding -a ./dolly-v2-3b/ -m ./dolly-v2-7b/ -n 5
+          ./build/samples/cpp/speculative_decoding_lm/speculative_decoding_lm ./dolly-v2-7b/ ./dolly-v2-3b/ "Alan Turing was a" > predictions_speculative.txt
+          ./build/samples/cpp/greedy_causal_lm/greedy_causal_lm ./dolly-v2-7b/ "Alan Turing was a" > predictions_greedy.txt
+          python -c "
+          with open('predictions_greedy.txt', 'r') as f:
+              predicted_greedy = f.readline()
+          with open('predictions_speculative.txt', 'r') as f:
+              predicted_speculative = f.readline()
+          assert predicted_greedy == predicted_speculative
+          "
+          echo "Alan Turing was a" passed
 
   cpp-prompt_lookup_decoding_lm-ubuntu:
     runs-on: ubuntu-20.04-16-cores

diff --git a/samples/CMakeLists.txt b/samples/CMakeLists.txt
@@ -30,6 +30,7 @@ install(DIRECTORY
             cpp/whisper_speech_recognition
             cpp/text2image
             cpp/lora_greedy_causal_lm
+            cpp/speculative_decoding_lm
         DESTINATION samples/cpp COMPONENT cpp_samples_genai)
 
 install(DIRECTORY
@@ -39,6 +40,7 @@ install(DIRECTORY
             python/multinomial_causal_lm
             python/visual_language_chat
             python/whisper_speech_recognition
+            # python/speculative_decoding_lm
             # python/text2image
         DESTINATION samples/python COMPONENT cpp_samples_genai
         USE_SOURCE_PERMISSIONS)
diff --git a/samples/cpp/continuous_batching_accuracy/CMakeLists.txt b/samples/cpp/continuous_batching_accuracy/CMakeLists.txt
@@ -28,3 +28,7 @@ find_package(OpenVINO REQUIRED COMPONENTS Runtime)
 set(TARGET_NAME continuous_batching_accuracy)
 add_executable(${TARGET_NAME} ${TARGET_NAME}.cpp)
 target_link_libraries(${TARGET_NAME} PRIVATE openvino::genai cxxopts::cxxopts)
+
+set(TARGET_NAME_CB continuous_batching_speculative_decoding)
+add_executable(${TARGET_NAME_CB} ${TARGET_NAME_CB}.cpp)
+target_link_libraries(${TARGET_NAME_CB} PRIVATE openvino::genai cxxopts::cxxopts)
diff --git a/...tinuous_batching_speculative_decoding.cpp → ...tinuous_batching_speculative_decoding.cpp b/...tinuous_batching_speculative_decoding.cpp → ...tinuous_batching_speculative_decoding.cpp
@@ -55,7 +55,6 @@ int main(int argc, char* argv[]) try {
     ("m,model", "Path to model and tokenizers base directory", cxxopts::value<std::string>()->default_value("."))
     ("a,draft_model", "Path to assisting model base directory", cxxopts::value<std::string>()->default_value("."))
     ("d,device", "Target device to run the model", cxxopts::value<std::string>()->default_value("CPU"))
-    ("use_prefix", "Whether to use a prefix or not", cxxopts::value<bool>()->default_value("false"))
     ("h,help", "Print usage");
 
     cxxopts::ParseResult result;
@@ -77,14 +76,6 @@ int main(int argc, char* argv[]) try {
     const std::string model_path = result["model"].as<std::string>();
     const std::string draft_model_path = result["draft_model"].as<std::string>();
     const std::string device = result["device"].as<std::string>();
-    const bool use_prefix = result["use_prefix"].as<bool>();
-
-    std::string prefix_str =
-        "You are an advanced language model designed to assist users by providing accurate, "
-        "relevant, and helpful information. Your responses should be accurate, concise, contextual, "
-        "respectful, and helpful. The request is: ";
-
-    // create dataset
 
     std::vector<std::string> prompt_examples = {
         "What is OpenVINO?",
@@ -117,15 +108,14 @@ int main(int argc, char* argv[]) try {
 
     ov::genai::SchedulerConfig scheduler_config;
     // batch size
-    scheduler_config.max_num_batched_tokens = use_prefix ? 256 : 32;
+    scheduler_config.max_num_batched_tokens = 32;
     // cache params
     scheduler_config.num_kv_blocks = 364;
     scheduler_config.block_size = get_default_block_size(device);
     // mode - vLLM or dynamic_split_fuse
     scheduler_config.dynamic_split_fuse = dynamic_split_fuse;
     // vLLM specific params
     scheduler_config.max_num_seqs = 2;
-    scheduler_config.enable_prefix_caching = use_prefix;
 
     ov::genai::ContinuousBatchingPipeline pipe(model_path, scheduler_config, device, {ov::genai::draft_model(draft_model_path, device)});
     std::vector<ov::genai::GenerationResult> generation_results = pipe.generate(prompts, generation_config);

diff --git a/samples/cpp/speculative_decoding_lm/CMakeLists.txt b/samples/cpp/speculative_decoding_lm/CMakeLists.txt
@@ -1,34 +1,23 @@
 # Copyright (C) 2024 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 
-# start of dependencies
-
-include(FetchContent)
-
-if(POLICY CMP0135)
-    cmake_policy(SET CMP0135 NEW)
-endif()
-
-FetchContent_Declare(cxxopts
-    URL https://github.com/jarro2783/cxxopts/archive/refs/tags/v3.1.1.tar.gz
-    URL_HASH SHA256=523175f792eb0ff04f9e653c90746c12655f10cb70f1d5e6d6d9491420298a08)
-FetchContent_MakeAvailable(cxxopts)
-
-if(NOT TARGET nlohmann_json)
-    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)
-endif()
-
-find_package(OpenVINO REQUIRED COMPONENTS Runtime)
-
-# end of dependencies
+find_package(OpenVINOGenAI REQUIRED
+    PATHS
+        "${CMAKE_BINARY_DIR}"  # Reuse the package from the build.
+        ${OpenVINO_DIR}  # GenAI may be installed alogside OpenVINO.
+    NO_CMAKE_FIND_ROOT_PATH
+)
 
 set(TARGET_NAME speculative_decoding_lm)
 add_executable(${TARGET_NAME} ${TARGET_NAME}.cpp)
-target_link_libraries(${TARGET_NAME} PRIVATE openvino::genai cxxopts::cxxopts)
+target_link_libraries(${TARGET_NAME} PRIVATE openvino::genai)
+
+set_target_properties(${TARGET_NAME} PROPERTIES
+    COMPILE_PDB_NAME ${TARGET_NAME} 
+    # Ensure out of box LC_RPATH on macOS with SIP
+    INSTALL_RPATH_USE_LINK_PATH ON)
 
-set(TARGET_NAME_CB continuous_batching_speculative_decoding)
-add_executable(${TARGET_NAME_CB} ${TARGET_NAME_CB}.cpp)
-target_link_libraries(${TARGET_NAME_CB} PRIVATE openvino::genai cxxopts::cxxopts)
+install(TARGETS ${TARGET_NAME} 
+        RUNTIME DESTINATION samples_bin/
+        COMPONENT samples_bin
+        EXCLUDE_FROM_ALL)
diff --git a/samples/cpp/speculative_decoding_lm/speculative_decoding_lm.cpp b/samples/cpp/speculative_decoding_lm/speculative_decoding_lm.cpp
@@ -2,108 +2,29 @@
 // SPDX-License-Identifier: Apache-2.0
 
 #include <openvino/openvino.hpp>
-#include <cxxopts.hpp>
 
 #include "openvino/genai/llm_pipeline.hpp"
 
-void print_generation_result(const std::vector<std::string>& texts, const std::vector<float>& log_probs) {
-    for (size_t output_id = 0; output_id < texts.size(); ++output_id) {
-        std::cout << "Answer " << output_id << " (" << log_probs[output_id] << ") : " << texts[output_id] << std::endl;
-    }
-}
-
-std::vector<ov::genai::GenerationConfig> get_spec_decoding_generation_config_examples() {
-
-    // sampling param for speulative decoding
-    ov::genai::GenerationConfig generation_config_greedy_constant = ov::genai::greedy();
-    {
-        generation_config_greedy_constant.num_assistant_tokens = 5;
-    }
-
-    ov::genai::GenerationConfig generation_config_multinomial_constant = ov::genai::multinomial();
-    {
-        generation_config_multinomial_constant.num_assistant_tokens = 5;
-        generation_config_multinomial_constant.num_return_sequences = 1;
-    }
-
-    ov::genai::GenerationConfig generation_config_greedy_dynamic = ov::genai::greedy();
-    {
-        generation_config_greedy_dynamic.assistant_confidence_threshold = 0.8f;
-    }
-
-    ov::genai::GenerationConfig generation_config_multinomial_dynamic = ov::genai::multinomial();
-    {
-        generation_config_multinomial_dynamic.assistant_confidence_threshold = 0.8f;
-    }
-
-    return {
-        generation_config_greedy_constant,
-        generation_config_multinomial_constant,
-        generation_config_greedy_dynamic,
-        generation_config_multinomial_dynamic,
-    };
-}
-
 int main(int argc, char* argv[]) try {
-    // Command line options
-
-    cxxopts::Options options("accuracy_sample", "Help command");
-
-    options.add_options()
-    ("n,num_prompts", "A number of prompts", cxxopts::value<size_t>()->default_value("1"))
-    ("dynamic_split_fuse", "Whether to use dynamic split-fuse or vLLM scheduling", cxxopts::value<bool>()->default_value("false"))
-    ("m,model", "Path to model and tokenizers base directory", cxxopts::value<std::string>()->default_value("."))
-    ("a,draft_model", "Path to assisting model base directory", cxxopts::value<std::string>()->default_value("."))
-    ("d,device", "Target device to run the model", cxxopts::value<std::string>()->default_value("CPU"))
-    ("use_prefix", "Whether to use a prefix or not", cxxopts::value<bool>()->default_value("false"))
-    ("h,help", "Print usage");
-
-    cxxopts::ParseResult result;
-    try {
-        result = options.parse(argc, argv);
-    } catch (const cxxopts::exceptions::exception& e) {
-        std::cout << e.what() << "\n\n";
-        std::cout << options.help() << std::endl;
-        return EXIT_FAILURE;
-    }
-
-    if (result.count("help")) {
-        std::cout << options.help() << std::endl;
-        return EXIT_SUCCESS;
-    }
-
-    const size_t num_prompts = result["num_prompts"].as<size_t>();
-    const bool dynamic_split_fuse = result["dynamic_split_fuse"].as<bool>();
-    const std::string model_path = result["model"].as<std::string>();
-    const std::string draft_model_path = result["draft_model"].as<std::string>();
-    const std::string device = result["device"].as<std::string>();
-    const bool use_prefix = result["use_prefix"].as<bool>();
-
-    std::string prefix_str =
-        "You are an advanced language model designed to assist users by providing accurate, "
-        "relevant, and helpful information. Your responses should be accurate, concise, contextual, "
-        "respectful, and helpful. The request is: ";
-
-    // create dataset
-
-    std::vector<std::string> prompt_examples = {
-        "What is OpenVINO?",
-        "How are you?",
-        "What is your name?",
-        "Tell me something about Canada",
-        "What is OpenVINO?",
-    };
-
-    auto generation_config = get_spec_decoding_generation_config_examples();
-    auto default_config_size = generation_config.size();
-    for (size_t i = default_config_size; i < num_prompts; ++i) {
-        generation_config.push_back(generation_config[i % default_config_size]);
+    if (4 != argc) {
+        throw std::runtime_error(std::string{"Usage: "} + argv[0] + " <MODEL_DIR> <DRAFT_MODEL_DIR> '<PROMPT>'");
     }
 
-    std::vector<std::string> prompts(num_prompts);
-    for (size_t i = 0; i < num_prompts; ++i) {
-        prompts[i] = prompt_examples[i % prompt_examples.size()];
-    }
+    ov::genai::GenerationConfig config;
+    config.max_new_tokens = 100;
+    // Speculative decoding generation parameters are mutually excluded
+    // add parameter to enable speculative decoding to generate `num_assistant_tokens` candidates by draft_model per iteration
+    config.num_assistant_tokens = 5;
+    // add parameter to enable speculative decoding to generate candidates by draft_model while candidate probability is higher than `assistant_confidence_threshold`
+    // config.assistant_confidence_threshold = 0.4
+
+    std::string main_model_path = argv[1];
+    std::string draft_model_path = argv[2];
+    std::string prompt = argv[3];
+
+    // User can run main and draft model on different devices.
+    // Please, set device for main model in `LLMPipeline` constructor and in in `ov::genai::draft_model` for draft.
+    std::string main_device = "CPU", draft_device = main_device;
 
     // Perform the inference
     auto get_default_block_size = [](const std::string& device) {
@@ -116,34 +37,21 @@ int main(int argc, char* argv[]) try {
     };
 
     ov::genai::SchedulerConfig scheduler_config;
-    // batch size
-    scheduler_config.max_num_batched_tokens = use_prefix ? 256 : 32;
-    // cache params
-    scheduler_config.num_kv_blocks = 364;
-    scheduler_config.block_size = get_default_block_size(device);
-    // mode - vLLM or dynamic_split_fuse
-    scheduler_config.dynamic_split_fuse = dynamic_split_fuse;
-    // vLLM specific params
-    scheduler_config.max_num_seqs = 2;
-    scheduler_config.enable_prefix_caching = use_prefix;
+    scheduler_config.cache_size = 5;
+    scheduler_config.block_size = get_default_block_size(main_device);
 
-    // It's possible to construct a Tokenizer from a different path.
-    // If the Tokenizer isn't specified, it's loaded from the same folder.
-    ov::genai::LLMPipeline pipe(model_path, device, ov::genai::draft_model(draft_model_path, device), ov::genai::scheduler_config(scheduler_config));
+    // Example to run main_model on GPU and draft_model on CPU:
+    // ov::genai::LLMPipeline pipe(main_model_path, "GPU", ov::genai::draft_model(draft_model_path, "CPU"), ov::genai::scheduler_config(scheduler_config));
+    ov::genai::LLMPipeline pipe(main_model_path, main_device, ov::genai::draft_model(draft_model_path, draft_device), ov::genai::scheduler_config(scheduler_config));
 
-    if (use_prefix) {
-        std::cout << "Running inference for prefix to compute the shared prompt's KV cache..." << std::endl;
-        auto generation_results = pipe.generate(prefix_str, ov::genai::greedy());
-    }
+    auto streamer = [](std::string subword) {
+        std::cout << subword << std::flush;
+        return false;
+    };
 
-    for (size_t request_id = 0; request_id < prompts.size(); ++request_id) {
-        ov::genai::DecodedResults generation_results = pipe.generate(prompts[request_id], generation_config[request_id]);
-        std::cout << "Question: " << prompts[request_id] << std::endl;
-        const std::vector<std::string>& text_results = generation_results.texts;
-        const std::vector<float>& log_prob_results = generation_results.scores;
-        print_generation_result(text_results, log_prob_results);
-        std::cout << std::endl;
-    }
+    // Since the streamer is set, the results will
+    // be printed each time a new token is generated.
+    pipe.generate(prompt, config, streamer);
 } catch (const std::exception& error) {
     try {
         std::cerr << error.what() << '\n';

diff --git a/src/cpp/include/openvino/genai/llm_pipeline.hpp b/src/cpp/include/openvino/genai/llm_pipeline.hpp
@@ -272,12 +272,35 @@ class OPENVINO_GENAI_EXPORTS LLMPipeline {
 OPENVINO_GENAI_EXPORTS std::pair<std::string, Any> streamer(StreamerVariant func);
 OPENVINO_GENAI_EXPORTS std::pair<std::string, Any> generation_config(const GenerationConfig& config);
 
-OPENVINO_GENAI_EXPORTS
-std::pair<std::string, Any> draft_model(
+OPENVINO_GENAI_EXPORTS std::pair<std::string, Any> _draft_model(
     const std::string& model_path,
-    const std::string& device = "",
-    const ov::AnyMap& plugin_config = {},
-    const ov::genai::SchedulerConfig& scheduler_config = {});
+    const std::string& device,
+    const ov::AnyMap& llm_config);
+
+template <typename... Properties,
+          typename std::enable_if<ov::util::StringAny<Properties...>::value, bool>::type = true>
+inline std::pair<std::string, Any> draft_model(
+    const std::string& model_path,
+    const std::string& device,
+    Properties&&... properties) {
+    return _draft_model(model_path, device, ov::AnyMap{std::forward<Properties>(properties)...});
+}
+
+template <typename... Properties,
+          typename std::enable_if<ov::util::StringAny<Properties...>::value, bool>::type = true>
+inline std::pair<std::string, Any> draft_model(
+    const std::string& model_path,
+    Properties&&... properties) {
+    return _draft_model(model_path, "", ov::AnyMap{std::forward<Properties>(properties)...});
+}
+
 
+inline std::pair<std::string, Any> 
+draft_model(
+    const std::string& model_path,
+    const std::string& device = "",
+    const ov::AnyMap& llm_config = ov::AnyMap()) {
+    return _draft_model(model_path, device, llm_config);
+}
 }  // namespace genai
 }  // namespace ov
diff --git a/src/cpp/src/continuous_batching_impl_interface.hpp b/src/cpp/src/continuous_batching_impl_interface.hpp
@@ -28,11 +28,11 @@ class ContinuousBatchingPipeline::ImplInterface {
         float m_infer_total_ms = 0.0f;
 
         ~PerfTime() {
-            std::cout << "Inference requests aggregated statistic: " << std::endl;
-            std::cout << "Paged attention % of inference execution: " << (m_paged_attention_time_ms / m_infer_total_ms) * 100 << std::endl;
-            std::cout << "MatMul % of inference execution: " << (m_matmul_time_ms / m_infer_total_ms) * 100 << std::endl;
-            std::cout << "Total inference execution secs: " << m_infer_total_ms / 1000. << std::endl;
-            std::cout << std::endl;
+            // std::cout << "Inference requests aggregated statistic: " << std::endl;
+            // std::cout << "Paged attention % of inference execution: " << (m_paged_attention_time_ms / m_infer_total_ms) * 100 << std::endl;
+            // std::cout << "MatMul % of inference execution: " << (m_matmul_time_ms / m_infer_total_ms) * 100 << std::endl;
+            // std::cout << "Total inference execution secs: " << m_infer_total_ms / 1000. << std::endl;
+            // std::cout << std::endl;
         }
     } m_perf;
     bool m_is_chat_conversation = false;

diff --git a/src/cpp/src/generation_config.cpp b/src/cpp/src/generation_config.cpp
@@ -165,9 +165,9 @@ void GenerationConfig::validate() const {
     }
     if (is_speculative_decoding()) {
         if (assistant_confidence_threshold != 0.f) {
-            OPENVINO_ASSERT(num_assistant_tokens == 0);
+            OPENVINO_ASSERT(num_assistant_tokens == 0, "Parameters `assistant_confidence_threshold` and `num_assistant_tokens` are mutually excluded in `GenerationConfig`");
         } else {
-            OPENVINO_ASSERT(num_assistant_tokens > 0);
+            OPENVINO_ASSERT(num_assistant_tokens > 0, "Parameters `assistant_confidence_threshold` and `num_assistant_tokens` are mutually excluded in `GenerationConfig`");
         };
     }
 }
@@ -202,5 +202,6 @@ GenerationConfig multinomial() {
     return multinomial_config;
 }
 
+
 }  // namespace genai
 }  // namespace ov