chat: fix build on Windows and Nomic Embed path on macOS (#2467)

* chat: remove unused oscompat source files These files are no longer needed now that the hnswlib index is gone. This fixes an issue with the Windows build as there was a compilation error in oscompat.cpp. Signed-off-by: Jared Van Bortel <jared@nomic.ai> * llm: fix pragma to be recognized by MSVC Replaces this MSVC warning: C:\msys64\home\Jared\gpt4all\gpt4all-chat\llm.cpp(53,21): warning C4081: expected '('; found 'string' With this: C:\msys64\home\Jared\gpt4all\gpt4all-chat\llm.cpp : warning : offline installer build will not check for updates! Signed-off-by: Jared Van Bortel <jared@nomic.ai> * usearch: fork usearch to fix `CreateFile` build error Signed-off-by: Jared Van Bortel <jared@nomic.ai> * dlhandle: fix incorrect assertion on Windows SetErrorMode returns the previous value of the error mode flags, not an indicator of success. Signed-off-by: Jared Van Bortel <jared@nomic.ai> * llamamodel: fix UB in LLamaModel::embedInternal It is undefined behavior to increment an STL iterator past the end of the container. Use offsets to do the math instead. Signed-off-by: Jared Van Bortel <jared@nomic.ai> * cmake: install embedding model to bundle's Resources dir on macOS Signed-off-by: Jared Van Bortel <jared@nomic.ai> * ci: fix macOS build by explicitly installing Rosetta Signed-off-by: Jared Van Bortel <jared@nomic.ai> --------- Signed-off-by: Jared Van Bortel <jared@nomic.ai>
2024-10-01 01:06:10 -04:00 · 2024-06-25 17:22:51 -04:00 · 2024-06-25 17:22:51 -04:00 · 88d85be0f9
commit 88d85be0f9
parent bbf0c2f246
10 changed files with 61 additions and 125 deletions
--- a/.circleci/continue_config.yml
+++ b/.circleci/continue_config.yml
@ -40,6 +40,9 @@ jobs:
      - restore_cache:  # this is the new step to restore cache
          keys:
            - macos-qt-cache-v3
+      - run:
+          name: Install Rosetta
+          command: softwareupdate --install-rosetta --agree-to-license  # needed for QtIFW
      - run:
          name: Installing Qt
          command: |
@ -325,6 +328,9 @@ jobs:
      - restore_cache:  # this is the new step to restore cache
          keys:
            - macos-qt-cache-v3
+      - run:
+          name: Install Rosetta
+          command: softwareupdate --install-rosetta --agree-to-license  # needed for QtIFW
      - run:
          name: Installing Qt
          command: |
--- a/.gitmodules
+++ b/.gitmodules
@ -4,4 +4,4 @@
 	branch = master
 [submodule "gpt4all-chat/usearch"]
 	path = gpt4all-chat/usearch
-	url = https://github.com/unum-cloud/usearch.git
+	url = https://github.com/nomic-ai/usearch.git
--- a/gpt4all-backend/dlhandle.cpp
+++ b/gpt4all-backend/dlhandle.cpp
@ -46,13 +46,11 @@ Dlhandle::Dlhandle(const fs::path &fpath)

    // Suppress the "Entry Point Not Found" dialog, caused by outdated nvcuda.dll from the GPU driver
    UINT lastErrorMode = GetErrorMode();
-    UINT success = SetErrorMode(lastErrorMode | SEM_FAILCRITICALERRORS);
-    assert(success);
+    SetErrorMode(lastErrorMode | SEM_FAILCRITICALERRORS);

    chandle = LoadLibraryExW(afpath.c_str(), NULL, LOAD_LIBRARY_SEARCH_DEFAULT_DIRS | LOAD_LIBRARY_SEARCH_DLL_LOAD_DIR);

-    success = SetErrorMode(lastErrorMode);
-    assert(success);
+    SetErrorMode(lastErrorMode);

    if (!chandle) {
        DWORD err = GetLastError();
--- a/gpt4all-backend/llamamodel.cpp
+++ b/gpt4all-backend/llamamodel.cpp
@ -1003,14 +1003,14 @@ void LLamaModel::embedInternal(
    size_t totalTokens = 0;
    for (unsigned i = 0; i < inputs.size(); i++) {
        auto &input = inputs[i];
-        for (auto it = input.begin(); it < input.end(); it += max_len) {
-            if (it > input.begin()) { it -= chunkOverlap; }
-            auto end = std::min(it + max_len, input.end());
+        for (unsigned j = 0; j < input.size(); j += max_len) {
+            if (j) { j -= chunkOverlap; }
+            unsigned end = std::min(j + max_len, unsigned(input.size()));
            batches.push_back({ i, {} });
            auto &batch = batches.back().batch;
            batch = prefixTokens;
-            batch.insert(batch.end(), it, end);
-            totalTokens += end - it;
+            batch.insert(batch.end(), input.begin() + j, input.begin() + end);
+            totalTokens += end - j;
            batch.push_back(eos_token);
            if (!doMean) { break; /* limit text to one chunk */ }
        }
--- a/gpt4all-chat/CMakeLists.txt
+++ b/gpt4all-chat/CMakeLists.txt
@ -64,14 +64,16 @@ set (CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)

 add_subdirectory(../gpt4all-backend llmodel)

-set(METAL_SHADER_FILE)
-if(${CMAKE_SYSTEM_NAME} MATCHES Darwin)
-    set(METAL_SHADER_FILE ${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/default.metallib)
+set(CHAT_EXE_RESOURCES)
+
+# Metal shader library
+if (APPLE)
+    list(APPEND CHAT_EXE_RESOURCES "${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/default.metallib")
 endif()

-set(APP_ICON_RESOURCE)
+# App icon
 if (WIN32)
-    set(APP_ICON_RESOURCE "${CMAKE_CURRENT_SOURCE_DIR}/resources/gpt4all.rc")
+    list(APPEND CHAT_EXE_RESOURCES "${CMAKE_CURRENT_SOURCE_DIR}/resources/gpt4all.rc")
 elseif (APPLE)
    # The MACOSX_BUNDLE_ICON_FILE variable is added to the Info.plist
    # generated by CMake. This variable contains the .icns file name,
@ -82,8 +84,24 @@ elseif (APPLE)
    set(APP_ICON_RESOURCE "${CMAKE_CURRENT_SOURCE_DIR}/resources/gpt4all.icns")
    set_source_files_properties(${APP_ICON_RESOURCE} PROPERTIES
        MACOSX_PACKAGE_LOCATION "Resources")
+    list(APPEND CHAT_EXE_RESOURCES "${APP_ICON_RESOURCE}")
 endif()

+# Embedding model
+set(LOCAL_EMBEDDING_MODEL "nomic-embed-text-v1.5.f16.gguf")
+set(LOCAL_EMBEDDING_MODEL_MD5 "a5401e7f7e46ed9fcaed5b60a281d547")
+set(LOCAL_EMBEDDING_MODEL_PATH "${CMAKE_BINARY_DIR}/resources/${LOCAL_EMBEDDING_MODEL}")
+set(LOCAL_EMBEDDING_MODEL_URL "https://gpt4all.io/models/gguf/${LOCAL_EMBEDDING_MODEL}")
+message(STATUS "Downloading embedding model from ${LOCAL_EMBEDDING_MODEL_URL} ...")
+file(DOWNLOAD
+    "${LOCAL_EMBEDDING_MODEL_URL}"
+    "${LOCAL_EMBEDDING_MODEL_PATH}"
+    EXPECTED_HASH "MD5=${LOCAL_EMBEDDING_MODEL_MD5}"
+)
+message(STATUS "Embedding model downloaded to ${LOCAL_EMBEDDING_MODEL_PATH}")
+if (APPLE)
+    list(APPEND CHAT_EXE_RESOURCES "${LOCAL_EMBEDDING_MODEL_PATH}")
+endif()

 qt_add_executable(chat
    main.cpp
@ -102,9 +120,8 @@ qt_add_executable(chat
    server.h server.cpp
    logger.h logger.cpp
    responsetext.h responsetext.cpp
-    oscompat.h oscompat.cpp
-    ${METAL_SHADER_FILE}
    ${APP_ICON_RESOURCE}
+    ${CHAT_EXE_RESOURCES}
 )

 qt_add_qml_module(chat
@ -197,23 +214,18 @@ qt_add_qml_module(chat
 )

 set_target_properties(chat PROPERTIES
-    MACOSX_BUNDLE_GUI_IDENTIFIER gpt4all
-    MACOSX_BUNDLE_BUNDLE_VERSION ${PROJECT_VERSION}
-    MACOSX_BUNDLE_SHORT_VERSION_STRING ${PROJECT_VERSION_MAJOR}.${PROJECT_VERSION_MINOR}
-    MACOSX_BUNDLE TRUE
    WIN32_EXECUTABLE TRUE
 )

-if(${CMAKE_SYSTEM_NAME} MATCHES Darwin)
+if (APPLE)
    set_target_properties(chat PROPERTIES
+        MACOSX_BUNDLE TRUE
+        MACOSX_BUNDLE_GUI_IDENTIFIER gpt4all
+        MACOSX_BUNDLE_BUNDLE_VERSION ${PROJECT_VERSION}
+        MACOSX_BUNDLE_SHORT_VERSION_STRING ${PROJECT_VERSION_MAJOR}.${PROJECT_VERSION_MINOR}
+        RESOURCE "${CHAT_EXE_RESOURCES}"
        OUTPUT_NAME gpt4all
    )
-endif()
-
-if(METAL_SHADER_FILE)
-    set_target_properties(chat PROPERTIES
-        RESOURCE ${METAL_SHADER_FILE}
-    )
    add_dependencies(chat ggml-metal)
 endif()

@ -237,17 +249,6 @@ target_link_libraries(chat
    PRIVATE llmodel)


-# -- extra resources --
-
-set(LOCAL_EMBEDDING_MODEL "nomic-embed-text-v1.5.f16.gguf")
-set(LOCAL_EMBEDDING_MODEL_MD5 "a5401e7f7e46ed9fcaed5b60a281d547")
-file(DOWNLOAD
-    "https://gpt4all.io/models/gguf/${LOCAL_EMBEDDING_MODEL}"
-    "${CMAKE_BINARY_DIR}/resources/${LOCAL_EMBEDDING_MODEL}"
-    EXPECTED_HASH "MD5=${LOCAL_EMBEDDING_MODEL_MD5}"
-)
-
-
 # -- install --

 set(COMPONENT_NAME_MAIN ${PROJECT_NAME})
@ -314,9 +315,11 @@ if (LLMODEL_CUDA)
    endif()
 endif()

+if (NOT APPLE)
    install(FILES "${CMAKE_BINARY_DIR}/resources/${LOCAL_EMBEDDING_MODEL}"
            DESTINATION resources
            COMPONENT ${COMPONENT_NAME_MAIN})
+endif()

 set(CPACK_GENERATOR "IFW")
 set(CPACK_VERBATIM_VARIABLES YES)
--- a/gpt4all-chat/embllm.cpp
+++ b/gpt4all-chat/embllm.cpp
@ -71,7 +71,13 @@ bool EmbeddingLLMWorker::loadModel()
        return true;
    }

-    QString filePath = u"%1/../resources/%2"_s.arg(QCoreApplication::applicationDirPath(), LOCAL_EMBEDDING_MODEL);
+#ifdef Q_OS_DARWIN
+    static const QString embPathFmt = u"%1/../Resources/%2"_s;
+#else
+    static const QString embPathFmt = u"%1/../resources/%2"_s;
+#endif
+
+    QString filePath = embPathFmt.arg(QCoreApplication::applicationDirPath(), LOCAL_EMBEDDING_MODEL);
    if (!QFileInfo::exists(filePath)) {
        qWarning() << "WARNING: Local embedding model not found";
        return false;
--- a/gpt4all-chat/llm.cpp
+++ b/gpt4all-chat/llm.cpp
@ -50,7 +50,7 @@ bool LLM::hasSettingsAccess() const
 bool LLM::checkForUpdates() const
 {
 #ifdef GPT4ALL_OFFLINE_INSTALLER
-    #pragma message "offline installer build will not check for updates!"
+#   pragma message(__FILE__ ": WARNING: offline installer build will not check for updates!")
    return QDesktopServices::openUrl(QUrl("https://gpt4all.io/"));
 #else
    Network::globalInstance()->trackEvent("check_for_updates");
--- a/gpt4all-chat/oscompat.cpp
+++ b/gpt4all-chat/oscompat.cpp
@ -1,70 +0,0 @@
-#include "oscompat.h"
-
-#include <QByteArray>
-#include <QString>
-#include <QtGlobal>
-
-#ifdef Q_OS_WIN32
-#   define WIN32_LEAN_AND_MEAN
-#   ifndef NOMINMAX
-#      define NOMINMAX
-#   endif
-#   include <windows.h>
-#   include <errno.h>
-#else
-#   include <fcntl.h>
-#   include <unistd.h>
-#endif
-
-bool gpt4all_fsync(int fd)
-{
-#if defined(Q_OS_WIN32)
-    HANDLE handle = HANDLE(_get_osfhandle(fd));
-    if (handle == INVALID_HANDLE_VALUE) {
-        errno = EBADF;
-        return false;
-    }
-
-    if (FlushFileBuffers(handle))
-        return true;
-
-    DWORD error = GetLastError();
-    switch (error) {
-    case ERROR_ACCESS_DENIED: // read-only file
-        return true;
-    case ERROR_INVALID_HANDLE: // not a regular file
-        errno = EINVAL;
-    default:
-        errno = EIO;
-    }
-
-    return false;
-#elif defined(Q_OS_DARWIN)
-    return fcntl(fd, F_FULLFSYNC, 0) == 0;
-#else
-    return fsync(fd) == 0;
-#endif
-}
-
-bool gpt4all_fdatasync(int fd)
-{
-#if defined(Q_OS_WIN32) || defined(Q_OS_DARWIN)
-    return gpt4all_fsync(fd);
-#else
-    return fdatasync(fd) == 0;
-#endif
-}
-
-bool gpt4all_syncdir(const QString &path)
-{
-#if defined(Q_OS_WIN32)
-    (void)path; // cannot sync a directory on Windows
-    return true;
-#else
-    int fd = open(path.toLocal8Bit().constData(), O_RDONLY | O_DIRECTORY);
-    if (fd == -1) return false;
-    bool ok = gpt4all_fdatasync(fd);
-    close(fd);
-    return ok;
-#endif
-}
--- a/gpt4all-chat/oscompat.h
+++ b/gpt4all-chat/oscompat.h
@ -1,7 +0,0 @@
-#pragma once
-
-class QString;
-
-bool gpt4all_fsync(int fd);
-bool gpt4all_fdatasync(int fd);
-bool gpt4all_syncdir(const QString &path);
--- a/gpt4all-chat/usearch
+++ b/gpt4all-chat/usearch
@ -1 +1 @@
-Subproject commit 5ea48c87c56a25ab57634a8f207f80ae675ed58e
+Subproject commit 22cfa3bd00ea542132ee826cdb220f9d6434bd43