文章目录
环境准备
YOLO V5 6.2
PyTorch Cuda
参考上文中 环境准备 YOLO V5 6.2 部分
准备代码
克隆仓库地址: git clone -b v6.2 https://github.com/ultralytics/yolov5.git
TensorRTx For YOLO V5 6.2
OpenCV
官网 OpenCV Release
GitHub OpenCV Release
opencv-4.6.0-vc14_vc15.exe
安装即解压, 到指定目录, 将 C:\mrathena\develop\opencv-4.6.0\build\x64\vc15\bin 配置到环境变量 Path
Cuda
安装选项 选 推荐的 精简 安装
安装时会自动将部分组件集成到 Visual Studio 开发环境中, 具体影响可以参考 这里, 我并不使用 VS, 所以选择打勾不安装
安装完成后, 执行 nvcc --version 或 nvcc -V, 如果输出如下信息, 则安装成功
TensorRT
执行 nvidia-smi, 驱动版本是 522.25, 可以支持到 CUDA 11.8. PyTorch 环境中 CUDA 选的是 11.6, 综合下来, TensorRT 也选支持 11.6 的
dirent.h
Dirent 是一个编程接口，用于检索有关 C 和 C++ 语言中的文件和目录的信息。该项目为 Microsoft Visual Studio 提供了一个 Dirent 接口。
dirent.h for windows, put into tensorrtx/include, download from https://github.com/tronkko/dirent
准备代码
克隆仓库地址: git clone -b yolov5-v6.2 https://github.com/wang-xinyu/tensorrtx.git
从 .pt 生成 .wts
拷贝 TensorRTx/yolov5/gen_wts.py 到 YOLOV5 环境中, 在 conda 激活 yolo 环境后执行转换命令, 生成 yolov5s.wts
cp {tensorrtx}/yolov5/gen_wts.py {ultralytics}/yolov5
cd {ultralytics}/yolov5
python gen_wts.py -w yolov5s.pt -o yolov5s.wts
// a file 'yolov5s.wts' will be generated.
构建并运行 TensorRTx/yolov5, C++
配置 CLion
CLion 开发 CUDA 应用, 工具链必须使用 Visual Studio. 安装 Visual Studio 时必须选择 C++ 桌面开发里的 MSVC 和 Windows SDK
初始 CMakeLists.txt
cmake_minimum_required(VERSION 2.6)
project(yolov5)
add_definitions(-std=c++11)
add_definitions(-DAPI_EXPORTS)
option(CUDA_USE_STATIC_CUDA_RUNTIME OFF)
set(CMAKE_CXX_STANDARD 11)
set(CMAKE_BUILD_TYPE Debug)
find_package(CUDA REQUIRED)
if(WIN32)
enable_language(CUDA)
endif(WIN32)
include_directories(${PROJECT_SOURCE_DIR}/include)
# include and link dirs of cuda and tensorrt, you need adapt them if yours are different
# cuda
include_directories(/usr/local/cuda/include)
link_directories(/usr/local/cuda/lib64)
# tensorrt
include_directories(/usr/include/x86_64-linux-gnu/)
link_directories(/usr/lib/x86_64-linux-gnu/)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -Wall -Ofast -g -Wfatal-errors -D_MWAITXINTRIN_H_INCLUDED")
cuda_add_library(myplugins SHARED yololayer.cu)
target_link_libraries(myplugins nvinfer cudart)
find_package(OpenCV)
include_directories(${OpenCV_INCLUDE_DIRS})
cuda_add_executable(yolov5 calibrator.cpp yolov5.cpp preprocess.cu)
target_link_libraries(yolov5 nvinfer)
target_link_libraries(yolov5 cudart)
target_link_libraries(yolov5 myplugins)
target_link_libraries(yolov5 ${OpenCV_LIBS})
add_executable(yolov5-cls calibrator.cpp yolov5_cls.cpp)
target_link_libraries(yolov5-cls nvinfer)
target_link_libraries(yolov5-cls cudart)
target_link_libraries(yolov5-cls myplugins)
target_link_libraries(yolov5-cls ${OpenCV_LIBS})
if(UNIX)
add_definitions(-O2 -pthread)
endif(UNIX)
修改 CMakeLists.txt 的过程
我使用的 IDE 是 CLion, 它也是使用该目录自带的 CMakeLists.txt 来构建工程, Build
警告
CMake Deprecation Warning at CMakeLists.txt:1 (cmake_minimum_required):
Compatibility with CMake < 2.8.12 will be removed from a future version of
CMake.
Update the VERSION argument <min> value or use a ...<max> suffix to tell
CMake that the project does not need compatibility with older versions.
意思是版本小于 2.8.12 的未来可能会移除 … 先改个 2.8.12
报错
CMake Error at C:/mrathena/develop/clion-2021.2.4/bin/cmake/win/share/cmake-3.20/Modules/FindCUDA.cmake:859 (message):
Specify CUDA_TOOLKIT_ROOT_DIR
Call Stack (most recent call first):
CMakeLists.txt:11 (find_package)
意思是 find_package 找不到 CUDA 目录, 需要我们指定 CUDA_TOOLKIT_ROOT_DIR, 在它上一行添加
set(CUDA_TOOLKIT_ROOT_DIR "C:\\Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.8")
又报错
CMake Error at C:/mrathena/develop/clion-2021.2.4/bin/cmake/win/share/cmake-3.20/Modules/FindCUDA.cmake:903 (find_program):
Syntax error in cmake code at
C:/mrathena/develop/clion-2021.2.4/bin/cmake/win/share/cmake-3.20/Modules/FindCUDA.cmake:903
when parsing string
CUDA_NVCC_EXECUTABLE;NAMES;nvcc;PATHS;C:\Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.8;ENV;CUDA_PATH;ENV;CUDA_BIN_PATH;PATH_SUFFIXES;bin;bin64;NO_DEFAULT_PATH
Invalid escape sequence \P
Call Stack (most recent call first):
C:/mrathena/develop/clion-2021.2.4/bin/cmake/win/share/cmake-3.20/Modules/FindCUDA.cmake:913 (cuda_find_host_program)
CMakeLists.txt:12 (find_package)
无效字符 \P, 路径要用 / 而不是 \ 或 \\
补充
根据 教程, 可以在 CMakeLists.txt 中打印一些信息
在 find_package(CUDA REQUIRED) 下添加
message(STATUS "libraries: ${CUDA_LIBRARIES}")
message(STATUS "include path: ${CUDA_INCLUDE_DIRS}")
可以打印出
-- libraries: C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.8/lib/x64/cudart.lib
-- include path: C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.8/include
报错
CMake Error at C:/mrathena/develop/clion-2021.2.4/bin/cmake/win/share/cmake-3.20/Modules/CMakeDetermineCUDACompiler.cmake:174 (message):
Failed to find nvcc.
Compiler requires the CUDA toolkit. Please set the CUDAToolkit_ROOT
variable.
Call Stack (most recent call first):
CMakeLists.txt:18 (enable_language)
找不到 nvcc, 指定 CUDAToolkit_ROOT
https://www.jetbrains.com/help/clion/cuda-projects.html#set-nvcc
https://www.jetbrains.com/help/clion/2021.2/quick-tutorial-on-configuring-clion-on-windows.html#MSVC
https://visualstudio.microsoft.com/zh-hans/vs/older-downloads/
C:\mrathena\develop\clion-2021.2.4\bin\cmake\win\bin\cmake.exe -DCMAKE_BUILD_TYPE=Debug -DCMAKE_DEPENDS_USE_COMPILER=FALSE -G "CodeBlocks - NMake Makefiles" C:\mrathena\develop\workspace\clion\test
-- The C compiler identification is MSVC 19.29.30146.0
-- The CXX compiler identification is MSVC 19.29.30146.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - failed
-- Check for working C compiler: C:/mrathena/develop/visual.studio-2019/VC/Tools/MSVC/14.29.30133/bin/Hostx64/x64/cl.exe
-- Check for working C compiler: C:/mrathena/develop/visual.studio-2019/VC/Tools/MSVC/14.29.30133/bin/Hostx64/x64/cl.exe - broken
CMake Error at C:/mrathena/develop/clion-2021.2.4/bin/cmake/win/share/cmake-3.20/Modules/CMakeTestCCompiler.cmake:66 (message):
The C compiler
"C:/mrathena/develop/visual.studio-2019/VC/Tools/MSVC/14.29.30133/bin/Hostx64/x64/cl.exe"
is not able to compile a simple test program.
It fails with the following output:
Change Dir: C:/mrathena/develop/workspace/clion/test/cmake-build-debug/CMakeFiles/CMakeTmp
Run Build Command(s):nmake -f Makefile /nologo cmTC_c73c3\fast && C:\mrathena\develop\visual.studio-2019\VC\Tools\MSVC\14.29.30133\bin\HostX64\x64\nmake.exe -f CMakeFiles\cmTC_c73c3.dir\build.make /nologo -L CMakeFiles\cmTC_c73c3.dir\build
Building C object CMakeFiles/cmTC_c73c3.dir/testCCompiler.c.obj
C:\mrathena\develop\clion-2021.2.4\bin\cmake\win\bin\cmake.exe -E cmake_cl_compile_depends --dep-file=CMakeFiles\cmTC_c73c3.dir\testCCompiler.c.obj.d --working-dir=C:\mrathena\develop\workspace\clion\test\cmake-build-debug\CMakeFiles\CMakeTmp --filter-prefix="" -- C:\mrathena\develop\visual.studio-2019\VC\Tools\MSVC\14.29.30133\bin\Hostx64\x64\cl.exe @C:\Users\mrathena\AppData\Local\Temp\nm85F9.tmp
Linking C executable cmTC_c73c3.exe
C:\mrathena\develop\clion-2021.2.4\bin\cmake\win\bin\cmake.exe -E vs_link_exe --intdir=CMakeFiles\cmTC_c73c3.dir --rc=rc --mt=CMAKE_MT-NOTFOUND --manifests -- C:\mrathena\develop\visual.studio-2019\VC\Tools\MSVC\14.29.30133\bin\Hostx64\x64\link.exe /nologo @CMakeFiles\cmTC_c73c3.dir\objects1.rsp @C:\Users\mrathena\AppData\Local\Temp\nm8697.tmp
RC Pass 1: command "rc /fo CMakeFiles\cmTC_c73c3.dir/manifest.res CMakeFiles\cmTC_c73c3.dir/manifest.rc" failed (exit code 0) with the following output:
系统找不到指定的文件。NMAKE : fatal error U1077: “C:\mrathena\develop\clion-2021.2.4\bin\cmake\win\bin\cmake.exe”: 返回代码“0xffffffff”
Stop.
NMAKE : fatal error U1077: “C:\mrathena\develop\visual.studio-2019\VC\Tools\MSVC\14.29.30133\bin\HostX64\x64\nmake.exe”: 返回代码“0x2”
Stop.
CMake will not be able to correctly generate this project.
Call Stack (most recent call first):
CMakeLists.txt:2 (project)
-- Configuring incomplete, errors occurred!
See also "C:/mrathena/develop/workspace/clion/test/cmake-build-debug/CMakeFiles/CMakeOutput.log".
See also "C:/mrathena/develop/workspace/clion/test/cmake-build-debug/CMakeFiles/CMakeError.log".
[Finished]