Merge branch 'main' into feature/model-lock

expose batch size flag to cli

.github/workflows/build-commit.yml

@@ -92,7 +92,7 @@ jobs:
         submodules: true
 
     - name: Install Dependencies
-      run: sudo apt-get update && sudo apt-get install -yq libboost-dev libasio-dev
+      run: sudo apt-get update && sudo apt-get install -yq libboost-dev libasio-dev libboost-thread-dev
 
     - name: Install OpenBlas
       if: ${{ matrix.build == 'avx2-openblas' }}
@@ -207,6 +207,7 @@ jobs:
           $msvc = $(join-path $vcdir $('VC\Tools\MSVC\'+$(gc -raw $(join-path $vcdir 'VC\Auxiliary\Build\Microsoft.VCToolsVersion.default.txt')).Trim()))
           $lib =  $(join-path $msvc 'bin\Hostx64\x64\lib.exe')
           & $lib /machine:x64 "/def:${env:RUNNER_TEMP}/openblas/lib/libopenblas.def" "/out:${env:RUNNER_TEMP}/openblas/lib/openblas.lib" /name:openblas.dll
+
       - name: Build
         id: cmake_build
         env:
@@ -214,7 +215,7 @@ jobs:
         run: |
           mkdir build
           cd build
-          cmake .. ${{ matrix.defines }}
+          cmake .. ${{ matrix.defines }} -DBoost_LIBRARYDIRS=${{ steps.install-boost.outputs.BOOST_ROOT }}/lib
           cmake --build . --config Release --target turbopilot
 
       # - name: Add libopenblas.dll

CMakeLists.txt

@@ -1,5 +1,15 @@
 cmake_minimum_required (VERSION 3.0)
 project(turbopilot VERSION 0.1.0)
+set(CMAKE_CXX_STANDARD 11)
+set(CMAKE_CXX_STANDARD_REQUIRED true)
+set(CMAKE_C_STANDARD 11)
+set(CMAKE_C_STANDARD_REQUIRED true)
+set(THREADS_PREFER_PTHREAD_FLAG ON)
+find_package(Threads REQUIRED)
+
+
+# option(BUILD_SHARED_LIBS "Build using shared libraries" OFF)
+
 
 set(CMAKE_EXPORT_COMPILE_COMMANDS "on")
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
@@ -27,6 +37,9 @@ if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm" OR ${CMAKE_SYSTEM_PROCESSOR} MATCHES
 endif()
 
 
+
+
+
 add_subdirectory(extern/ggml)
 add_subdirectory(extern/argparse)
 add_subdirectory(extern/spdlog)
@@ -35,8 +48,11 @@ if (GGML_STATIC)
     SET(CMAKE_FIND_LIBRARY_SUFFIXES ".a")
     SET(BUILD_SHARED_LIBS OFF)
     SET(CMAKE_EXE_LINKER_FLAGS "-static")
-endif()
 
+    # if(GGML_OPENBLAS)
+    #     set(BLA_STATIC ON)
+    # endif()
+endif()
 
 
 add_subdirectory(src)

Dockerfile.cuda11

@@ -7,7 +7,7 @@ RUN apt-get update && apt-get install ca-certificates gpg wget
 RUN wget -O - https://apt.kitware.com/keys/kitware-archive-latest.asc 2>/dev/null | gpg --dearmor - | tee /usr/share/keyrings/kitware-archive-keyring.gpg >/dev/null
 RUN echo 'deb [signed-by=/usr/share/keyrings/kitware-archive-keyring.gpg] https://apt.kitware.com/ubuntu/ jammy main' | tee /etc/apt/sources.list.d/kitware.list >/dev/null
 
-RUN apt-get update && apt-get install -y build-essential cmake libboost-dev
+RUN apt-get update && apt-get install -y build-essential cmake libboost-dev libboost-thread-dev
 
 
 ADD ./ /turbopilot

Dockerfile.cuda12

@@ -7,7 +7,7 @@ RUN apt-get update && apt-get install ca-certificates gpg wget
 RUN wget -O - https://apt.kitware.com/keys/kitware-archive-latest.asc 2>/dev/null | gpg --dearmor - | tee /usr/share/keyrings/kitware-archive-keyring.gpg >/dev/null
 RUN echo 'deb [signed-by=/usr/share/keyrings/kitware-archive-keyring.gpg] https://apt.kitware.com/ubuntu/ focal main' | tee /etc/apt/sources.list.d/kitware.list >/dev/null
 
-RUN apt-get update && apt-get install -y build-essential cmake libboost-dev
+RUN apt-get update && apt-get install -y build-essential cmake libboost-dev libboost-thread-dev
 
 ADD ./ /turbopilot
 

Dockerfile.default

@@ -1,6 +1,6 @@
 FROM alpine AS build
 
-RUN apk add --update alpine-sdk boost-dev cmake asio-dev
+RUN apk add --update alpine-sdk boost-dev cmake asio-dev 
 
 ADD ./ /turbopilot/
 

MODELS.md

@@ -6,9 +6,9 @@
 
 | Model Name          | RAM Requirement | Direct Download  | HF Project Link |
 |---------------------|-----------------|-----------------|-----------------|
-| StarCoder   | ~3GiB        |   [:arrow_down:](https://huggingface.co/TheBloke/stablecode-instruct-alpha-3b-GGML/blob/main/stablecode-instruct-alpha-3b.ggmlv1.q4_0.bin)           |   [:hugs:](https://huggingface.co/TheBloke/stablecode-instruct-alpha-3b-GGML/)           |
-
+| StableCode   | ~3GiB        |   [:arrow_down:](https://huggingface.co/TheBloke/stablecode-instruct-alpha-3b-GGML/blob/main/stablecode-instruct-alpha-3b.ggmlv1.q4_0.bin)           |   [:hugs:](https://huggingface.co/TheBloke/stablecode-instruct-alpha-3b-GGML/)           |
 
+To run in Turbopilot set model type `-m stablecode`
 
 ## "Coder" family models
 
@@ -23,7 +23,7 @@ This model is primarily trained on Python, Java and Javscript.
 
 | Model Name          | RAM Requirement | Direct Download  | HF Project Link |
 |---------------------|-----------------|-----------------|-----------------|
-| StarCoder   | ~2GiB        |   [:arrow_down:](https://huggingface.co/mike-ravkine/gpt_bigcode-santacoder-GGML/resolve/main/santacoder-q4_0.bin)           |   [:hugs:](https://huggingface.co/mike-ravkine/gpt_bigcode-santacoder-GGML/)           |
+| SantaCoder   | ~2GiB        |   [:arrow_down:](https://huggingface.co/mike-ravkine/gpt_bigcode-santacoder-GGML/resolve/main/santacoder-q4_0.bin)           |   [:hugs:](https://huggingface.co/mike-ravkine/gpt_bigcode-santacoder-GGML/)           |
 
 To run in Turbopilot set model type `-m starcoder`
 
@@ -39,7 +39,7 @@ Even when quantized, WizardCoder is a large model that takes up a significant am
 |---------------------|-----------------|-----------------|-----------------|
 | WizardCoder   | ~12GiB        |   [:arrow_down:](https://huggingface.co/TheBloke/WizardCoder-15B-1.0-GGML/resolve/main/WizardCoder-15B-1.0.ggmlv3.q4_0.bin)           |   [:hugs:](https://huggingface.co/TheBloke/WizardCoder-15B-1.0-GGML/)           |
 
-To run in Turbopilot set model type `-m starcoder`
+To run in Turbopilot set model type `-m wizardcoder`
 
 
 ### StarCoder (Released 4/5/2023)

include/turbopilot/model.hpp

@@ -7,7 +7,7 @@
 #include <map>
 #include <vector>
 #include <random>
-#include <boost/thread.hpp>
+#include <mutex>
 
 typedef void (*offload_func_t)(struct ggml_tensor * tensor);
 void ggml_nop(struct ggml_tensor * tensor);
@@ -64,7 +64,7 @@ protected:
     virtual std::stringstream predict_impl(std::string prompt, int max_length, bool include_prompt) = 0;
     ModelConfig config;
     std::mt19937 &rng;
-    boost::mutex model_lock;
+    std::mutex model_lock;
 };
 
 #endif //__TURBOPILOT_MODEL_H

src/CMakeLists.txt

@@ -1,6 +1,7 @@
 set(TURBOPILOT_TARGET turbopilot)
 
-find_package(Boost REQUIRED)
+find_package(Boost COMPONENTS thread system REQUIRED)
+
 include_directories(${Boost_INCLUDE_DIRS})
 
 add_executable(${TURBOPILOT_TARGET} 
@@ -16,6 +17,9 @@ add_executable(${TURBOPILOT_TARGET}
   ../include/turbopilot/starcoder.hpp
   )
 
+#set(THREADS_PREFER_PTHREAD_FLAG TRUE)
+#find_package(Threads REQUIRED)
+
 
 target_include_directories(${TURBOPILOT_TARGET} PRIVATE
   ../include
@@ -23,8 +27,6 @@ target_include_directories(${TURBOPILOT_TARGET} PRIVATE
   ../extern/crow/include
 )
 
+#target_compile_features(${TURBOPILOT_TARGET} PRIVATE cxx_std_11)
 
-target_link_libraries(${TURBOPILOT_TARGET} PRIVATE ggml argparse)
-
-
-#target_link_libraries(${TURBOPILOT_TARGET} PRIVATE spdlog::spdlog_header_only)
+target_link_libraries(${TURBOPILOT_TARGET} PRIVATE ggml argparse)

src/main.cpp

@@ -60,6 +60,11 @@ int main(int argc, char **argv)
         .default_value(0.1f)
         .scan<'g', float>();
 
+    program.add_argument("-b", "--batch-size")
+        .help("set batch size for model completion")
+        .default_value(512)
+        .scan<'i',int>();
+
 
     program.add_argument("prompt").remaining();
 
@@ -96,6 +101,7 @@ int main(int argc, char **argv)
     config.n_threads = program.get<int>("--threads");
     config.temp = program.get<float>("--temperature");
     config.top_p = program.get<float>("--top-p");
+    config.n_batch = program.get<int>("--batch-size");
 
     if(model_type.compare("codegen") == 0) {
         spdlog::info("Initializing GPT-J type model for '{}' model", model_type);
@@ -136,6 +142,7 @@ int main(int argc, char **argv)
         return "Hello world";
     });
 
+
     CROW_ROUTE(app, "/copilot_internal/v2/token")([](){
         //return "Hello world";