Darknet CUDA/CUDANN环境的快速安装

原创 已于 2022-12-02 17:52:02 修改 · 置顶 · 4.3k 阅读 · 3 ·
CC 4.0 BY-SA版权
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
文章标签：

#深度学习 #pytorch #神经网络 #1024程序员节

于 2022-01-31 10:42:14 首次发布
在Ubuntu18.04系统上，通过安装Nvidia驱动、CUDA环境、Darknet并配置环境变量，详细步骤包括：验证CUDA环境、下载Darknet、安装OpenCV和Python接口、编译Darknet开启GPU支持、运行YOLOV3模型。使用nvidia-smi监控显卡状态，确保GPU加速提升了检测帧率。
部署运行你感兴趣的模型镜像

在具备GPU显卡且主持CUDA的纯净的UBUNTU18.04系统上，按照如下指导文档安装 Nvidia 显卡驱动。

Ubuntu18.04安装CUDA深度学习环境_tugouxp的专栏-优快云博客

之后，下载darknet

git clone https://github.com/AlexeyAB/darknet.git

进入darknet/scripts，执行命令：

sudo ./scripts/setup.sh --InstallCUDA

最后输出pwsh的错误 貌似没有太大影响。

在.bashrc文件中添加如下设置环境变量：

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-11.5/lib64
export PATH=$PATH:/usr/local/cuda-11.5/bin:/usr/local/cuda/bin
export CUDA_HOME=$CUDA_HOME:/usr/local/cuda-11.5

安装opencv, python opencv:

sudo apt install libopencv-dev

之后安装Anaconda3-2020.11-Linux-x86_64.sh参考下面博客

Cuda环境搭建并部署多目标跟踪模型FairMOT推理验证_tugouxp的专栏-优快云博客

安装完毕anaconda后，升级PIP工具，不升级会在安装PYTHON OPENCV的时候失败。

pip3 install --upgrade pip

之后执行

pip3 install -i https://pypi.tuna.tsinghua.edu.cn/simple scikit-build
pip3 install opencv-python

编译darknet,打开GPU，CUDA,OPENCV支持。

下载YOLOV3权重

wget -c https://pjreddie.com/media/files/yolov3.weights

验证单张图片，执行./image_yolov3.sh

验证摄像头采集

修改darknet_video.py实现，将默认的YOLOV4模型修改为YOLOV3模型。

最后，执行python3 darknet_video.py

或者不修改原始测试文件，直接执行

 python darknet_video.py --config_file=./cfg/yolov3.cfg --weights=yolov3.weights

由于使能了GPU，检测帧率有明显提升。输入命令watch -n 1 nvidia-smi命令监视显卡一举一动：

使用显卡的有三个程序，由于我们将nouveau加入了黑名单，所以XORG使用了NVIDIA的显卡驱动做渲染。同时我开了一个python作实时的码流检测。同时开启了CUDA-Z查看显卡信息。

补充：

中科大和ALIYUN的下载源，替换后不是一般的快,但是这样作也有一个缺点，就是用apt命令安装某些软件包的时候，会发生系统认为当前的版本为非LTS支持版本而无法处理依赖，导致安装进行不下去的情况，则个时候需要恢复原来的source.list文件中的内容，才能恢复。

deb http://mirrors.ustc.edu.cn/ubuntu/ xenial main restricted universe multiverse
deb http://mirrors.ustc.edu.cn/ubuntu/ xenial-security main restricted universe multiverse
deb http://mirrors.ustc.edu.cn/ubuntu/ xenial-updates main restricted universe multiverse
deb http://mirrors.ustc.edu.cn/ubuntu/ xenial-proposed main restricted universe multiverse
deb http://mirrors.ustc.edu.cn/ubuntu/ xenial-backports main restricted universe multiverse
deb-src http://mirrors.ustc.edu.cn/ubuntu/ xenial main restricted universe multiverse
deb-src http://mirrors.ustc.edu.cn/ubuntu/ xenial-security main restricted universe multiverse
deb-src http://mirrors.ustc.edu.cn/ubuntu/ xenial-updates main restricted universe multiverse
deb-src http://mirrors.ustc.edu.cn/ubuntu/ xenial-proposed main restricted universe multiverse
deb-src http://mirrors.ustc.edu.cn/ubuntu/ xenial-backports main restricted universe multiverse
deb http://mirrors.aliyun.com/ubuntu/ xenial main
deb-src http://mirrors.aliyun.com/ubuntu/ xenial main
deb http://mirrors.aliyun.com/ubuntu/ xenial-updates main
deb-src http://mirrors.aliyun.com/ubuntu/ xenial-updates main
deb http://mirrors.aliyun.com/ubuntu/ xenial universe
deb-src http://mirrors.aliyun.com/ubuntu/ xenial universe
deb http://mirrors.aliyun.com/ubuntu/ xenial-updates universe
deb-src http://mirrors.aliyun.com/ubuntu/ xenial-updates universe
deb http://mirrors.aliyun.com/ubuntu/ xenial-security main
deb-src http://mirrors.aliyun.com/ubuntu/ xenial-security main
deb http://mirrors.aliyun.com/ubuntu/ xenial-security universe
deb-src http://mirrors.aliyun.com/ubuntu/ xenial-security universe

总结：

从验证结果看，darknet环境安装--InstallCUDA并不会自动安装N卡驱动，为确保万无一失，N卡驱动需要单独安装，并且先于CUDA环境的安装。

GPU显卡驱动分为两类，一类叫做UMD（user mode driver)和KMD（kernel mode driver).

nvidia.ko属于KMD，KMD是真正跟GPU打交道的Driver， 而opengl,libcuda.so, libcudart.so则属于UMD

无论是 nvidia.ko，还是 libcuda.so，还是 libcudart.so，都是被剥离了符号表的,大多数函数名是加密替换了的,nvidia的生态还是比较封闭的。

nvidia-smi principle:

strace -e trace=open,openat,read,write nvidia-smi
czl@czl-RedmiBook-14:~$ strace -e trace=open,openat,read,write nvidia-smi
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v3/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v2/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/x86_64/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/x86_64/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/x86_64/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "x86_64/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/glibc-hwcaps/x86-64-v3/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/glibc-hwcaps/x86-64-v2/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/tls/haswell/x86_64/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/tls/haswell/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/tls/x86_64/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/tls/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/haswell/x86_64/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/haswell/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/x86_64/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/libpthread.so.0", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/etc/ld.so.cache", O_RDONLY|O_CLOEXEC) = 3
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libpthread.so.0", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\0\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\0\0\0\0\0\0\0\0"..., 832) = 832
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v3/libm.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v2/libm.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/x86_64/libm.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/libm.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/x86_64/libm.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/libm.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/x86_64/libm.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/libm.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "x86_64/libm.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "libm.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/libm.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libm.so.6", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\3\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\0\0\0\0\0\0\0\0"..., 832) = 832
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v3/libdl.so.2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v2/libdl.so.2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/x86_64/libdl.so.2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/libdl.so.2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/x86_64/libdl.so.2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/libdl.so.2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/x86_64/libdl.so.2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/libdl.so.2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "x86_64/libdl.so.2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "libdl.so.2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/libdl.so.2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libdl.so.2", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\0\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\0\0\0\0\0\0\0\0"..., 832) = 832
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v3/libc.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v2/libc.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/x86_64/libc.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/libc.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/x86_64/libc.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/libc.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/x86_64/libc.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/libc.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "x86_64/libc.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "libc.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/libc.so.6", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libc.so.6", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\3\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0P\237\2\0\0\0\0\0"..., 832) = 832
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v3/librt.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v2/librt.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/x86_64/librt.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/librt.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/x86_64/librt.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/librt.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/x86_64/librt.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/librt.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "x86_64/librt.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "librt.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/librt.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/librt.so.1", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\0\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\0\0\0\0\0\0\0\0"..., 832) = 832
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v3/libnvidia-ml.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v2/libnvidia-ml.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/x86_64/libnvidia-ml.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/libnvidia-ml.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/x86_64/libnvidia-ml.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/libnvidia-ml.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/x86_64/libnvidia-ml.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/libnvidia-ml.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "x86_64/libnvidia-ml.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "libnvidia-ml.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/libnvidia-ml.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/etc/ld.so.cache", O_RDONLY|O_CLOEXEC) = 3
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libnvidia-ml.so.1", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\0\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\00000\1\0\0\0\0\0"..., 832) = 832
openat(AT_FDCWD, "/proc/modules", O_RDONLY) = 3
read(3, "binfmt_misc 24576 1 - Live 0x000"..., 1024) = 1024
read(3, "00000000000\nalgif_hash 16384 1 -"..., 1024) = 1024
read(3, "e 0x0000000000000000\nghash_clmul"..., 1024) = 1024
openat(AT_FDCWD, "/proc/driver/nvidia/params", O_RDONLY) = 3
read(3, "ResmanDebugLevel: 4294967295\nRmL"..., 1024) = 878
openat(AT_FDCWD, "/dev/nvidiactl", O_RDWR) = 3
openat(AT_FDCWD, "/sys/devices/system/memory/block_size_bytes", O_RDONLY) = 4
read(4, "8000000\n", 99)                = 8
openat(AT_FDCWD, "/proc/devices", O_RDONLY) = 4
read(4, "Character devices:\n  1 mem\n  4 /"..., 1024) = 779
openat(AT_FDCWD, "/proc/driver/nvidia/capabilities/mig/config", O_RDONLY) = 4
read(4, "DeviceFileMinor: 1\nDeviceFileMod"..., 1024) = 59
openat(AT_FDCWD, "/proc/driver/nvidia/capabilities/mig/config", O_RDONLY) = 4
read(4, "DeviceFileMinor: 1\nDeviceFileMod"..., 1024) = 59
read(4, "", 1024)                       = 0
openat(AT_FDCWD, "/proc/devices", O_RDONLY) = 4
read(4, "Character devices:\n  1 mem\n  4 /"..., 1024) = 779
openat(AT_FDCWD, "/proc/driver/nvidia/capabilities/mig/monitor", O_RDONLY) = 4
read(4, "DeviceFileMinor: 2\nDeviceFileMod"..., 1024) = 59
openat(AT_FDCWD, "/proc/driver/nvidia/capabilities/mig/monitor", O_RDONLY) = 4
read(4, "DeviceFileMinor: 2\nDeviceFileMod"..., 1024) = 59
read(4, "", 1024)                       = 0
openat(AT_FDCWD, "/dev/nvidia-caps/nvidia-cap2", O_RDONLY|O_CLOEXEC) = 4
openat(AT_FDCWD, "/proc/driver/nvidia/params", O_RDONLY) = 4
read(4, "ResmanDebugLevel: 4294967295\nRmL"..., 1024) = 878
openat(AT_FDCWD, "/dev/nvidia0", O_RDWR|O_CLOEXEC) = 4
openat(AT_FDCWD, "/proc/driver/nvidia/params", O_RDONLY) = 6
read(6, "ResmanDebugLevel: 4294967295\nRmL"..., 1024) = 878
openat(AT_FDCWD, "/dev/nvidia0", O_RDWR|O_CLOEXEC) = 6
openat(AT_FDCWD, "/proc/driver/nvidia/params", O_RDONLY) = 7
read(7, "ResmanDebugLevel: 4294967295\nRmL"..., 1024) = 878
openat(AT_FDCWD, "/dev/nvidia0", O_RDWR|O_CLOEXEC) = 7
openat(AT_FDCWD, "/etc/localtime", O_RDONLY|O_CLOEXEC) = 8
read(8, "TZif2\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 561
read(8, "TZif2\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 342
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v3/libcuda.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "glibc-hwcaps/x86-64-v2/libcuda.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/x86_64/libcuda.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/haswell/libcuda.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/x86_64/libcuda.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "tls/libcuda.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/x86_64/libcuda.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "haswell/libcuda.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "x86_64/libcuda.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "libcuda.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/local/cuda-11.7/lib64/libcuda.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/etc/ld.so.cache", O_RDONLY|O_CLOEXEC) = 8
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libcuda.so.1", O_RDONLY|O_CLOEXEC) = 8
read(8, "\177ELF\2\1\1\0\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\320M\r\0\0\0\0\0"..., 832) = 832
write(1, "Mon Oct 24 22:20:30 2022       \n", 32Mon Oct 24 22:20:30 2022       
) = 32
write(1, "+-------------------------------"..., 80+-----------------------------------------------------------------------------+
) = 80
write(1, "| NVIDIA-SMI 515.65.01    Driver"..., 80| NVIDIA-SMI 515.65.01    Driver Version: 515.65.01    CUDA Version: 11.7     |
) = 80
write(1, "|-------------------------------"..., 80|-------------------------------+----------------------+----------------------+
) = 80
write(1, "| GPU  Name        Persistence-M"..., 80| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
) = 80
write(1, "| Fan  Temp  Perf  Pwr:Usage/Cap"..., 80| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
) = 80
write(1, "|                               "..., 80|                               |                      |               MIG M. |
) = 80
write(1, "|==============================="..., 80|===============================+======================+======================|
) = 80
write(8, "\200\0\08\220H\273\253\0\0\0\0\0\0\0\2\0\0\210\276\0\0\0\1\0\0\0\2\0\0\0\0"..., 60) = 60
read(8, "\200\0\0 \220H\273\253\0\0\0\1\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 9000) = 36
write(1, "|   0  NVIDIA GeForce ...  Off  "..., 240|   0  NVIDIA GeForce ...  Off  | 00000000:02:00.0 Off |                  N/A |
| N/A   40C    P8    N/A /  N/A |      4MiB /  2048MiB |      0%      Default |
|                               |                      |                  N/A |
) = 240
write(1, "+-------------------------------"..., 80+-------------------------------+----------------------+----------------------+
) = 80
write(1, "                                "..., 480                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
) = 480
openat(AT_FDCWD, "/proc/2249/cmdline", O_RDONLY) = 8
read(8, "/usr/lib/xorg/Xorg\0vt2\0-displayf"..., 4096) = 144
read(8, "", 3072)                       = 0
write(1, "|    0   N/A  N/A      2249     "..., 80|    0   N/A  N/A      2249      G   /usr/lib/xorg/Xorg                  4MiB |
) = 80
write(1, "+-------------------------------"..., 80+-----------------------------------------------------------------------------+
) = 80
+++ exited with 0 +++
czl@czl-RedmiBook-14:~$

according the trace of nvida-smi, we can find the device node of the progra need open is:

NVIDIA显卡也是字符设备，BIREN GPGPU以及AMD也是如此，但并不代表所有的都是这样，有些显卡可能设备节点建立在DRM模块中，具体可以参考

AMD HSA 异构计算架构和AMD-KFD内核驱动_papaofdoudou的博客-优快云博客

结束！

您可能感兴趣的与本文相关的镜像

Python3.9

Python3.9

Conda
Python

Python 是一种高级、解释型、通用的编程语言，以其简洁易读的语法而闻名，适用于广泛的应用，包括Web开发、数据分析、人工智能和自动化脚本

cuda、cudann、nvcc、pytorch、torchvision
北落师门XY的博客
12-24 7130
一、定义 1.1、什么是CUDA CUDA(ComputeUnified Device Architecture)，是显卡厂商NVIDIA推出的运算平台。 CUDA是一种由NVIDIA推出的通用并行计算架构，该架构使GPU能够解决复杂的计算问题。 1.2、什么是CUDNN NVIDIA CUDNN是用于深度神经网络的GPU加速库。 二、查看版本 2.1 查看显卡驱动 nvidia-smi Driver Version: 410.104 CUDA Version: 10.0 表明显卡驱
NVIDIA+CUDA+cudaNN的配置与Anaconda虚拟环境的搭建--深度学习第一步
人工智能视觉分析算法学习实践和经验分享。
02-17 4258
目标： 从头开始，一步一步搭建深度学习环境 实现适用于python语言的深度学习网络训练环境 具体内容包括： 1.英伟达显卡安装和cuda配置安装； 2.虚拟环境Anaconda3的搭建； 3.测试实现torch1.4+gpu。 NVIDIA显卡的cuda和cuDNN配置： 提示：这里以win10下的NVIDIA GeForce RTX 3060为例，进行安装描述。 首先：找查看显卡信息的地方，NVIDIA Control Panel,因为需要知道下载那个版本的cuda来与之匹配。 查找版本信息：472
CUDA和cuDNN安装
最新发布
Dneccc的博客
10-24 660
CUDA+cuDNN安装极简教程
yolov4训练数据: cuda和cudnn的安装
Tjmies的博客
11-29 1551
yolov4训练数据: cuda和cudnn的安装
CUDA CUDNN安装篇
wang_rt的博客
07-11 909
包括疑问和一些奇怪的点！欢迎大家讨论！
安装cudann教程
junjian Li
08-10 3034
1.首先安装好cuda。 2.下载cudnn的安装，我这里用到的是cudnn v7.0版本， 3.将下载下来的tar文件进行解压： $ tar -xzvf cudnn-9.0-linux-x64-v7.tgz 4.接下来，在命令终端执行如下操作： $ cp cuda/include/cudnn.h /usr/local/cuda/include $ cp cuda/lib64/libcudn...
【Ubuntu｜cuDNN】训练 DarkNet 时解决 fatal error: cudnn_version.h
【人工智障 | 计算鸡视觉 | 机器孩】
08-07 1379
文章目录前言解决方法 前言 在训练 darknet yolo 时碰到这个问题，这是由于新旧版本的 cudann 中的文件有区别，新版本多了一些文件。 解决方法 回到之前安装 cudnn 的那篇博客，找到了自己当初安装的 cudnn 的版本。 进入官网：https://developer.nvidia.com/rdp/cudnn-archive，下载自己之前安装的版本（需要登陆）。 解压后，进入文件夹，执行：# 参考来源：https://github.com/pjreddie/darknet/issue
NVIDIA-CUDA HPC 编程模型与内存管理初探
tugouxp的专栏
11-12 3639
主存和现存在PCIE框架下可以做到互相访问，路径归纳如下：三种角色，两种存储，六种路径。1.HBM通过BAR透给了HOST，使HOST可以通过MMU给CPU访问，或者通过IOMMU给设备访问。2.相反，Host Memory并没有类似的BAR机制透给 PCIE设备端，PCIE设备端想要访问 HOST MEMORY必须经过 HOST IOMMU的映射一遍，加上PCIE端的SMMU映射，对于PCIE设备端（GPU）来说，要经过两层IOMMU翻译。所以从HOST端和GPU端看起来不太对称。
pjreddie 的 darknet , yolo 的 NodeJS 包装器 .zip
11-26
暗网.JSpjreddie 开源神经网络框架 Darknet 的 Node 包装器，使用外部函数接口库。阅读JavaScript 中的 YOLOv3。先决条件Linux、Windows（Linux 子系统）、节点构建工具...CUDA 和/或 CUDANN，请导出DARKNET_BUILD_WITH
ubuntu18安装、测试YOLOV3记录
qq_38988221的博客
12-28 1587
YOLOV3搬运工，记录本人安装测试过程
适用于异构芯片(CPU,ASIC,DSP,FPGA,GPU/GPGPU)的软件并行技术
tugouxp的专栏
11-14 7152
最后，这么多的并行计算方式，各有千秋，犹如八仙过海，各显神通，在哪里可以一睹他们的风采？无疑机器学习领域是其主战场。
cuda cudann pytorch安装.pdf
04-11
cuda cudann pytorch安装
【学习心得】安装cuda/cudann和pytorch
qq_39780701的博客
08-14 4301
本文介绍如何下载和安装pytorch和Cuda/Cudann
安装CUDA和CudaNN
weixin_50742178的博客
10-27 2533
把CUDA选一下，还需要注意一点，如果 你前面没有安装vs，直接安装的这个，需要把CUDA里面的Visual Studio Integration取消勾选，否则会安装不成功。1.找到下载好的cuda的安装包，双击打开，第一个是临时放置文件的路径，安装完成后会自动删除（随便设置到不含中文的路径下）在CUDA的安装路径（在CUDA安装过程中，我们保存过）出现了你安装的CUDA的版本信息，说明安装成功了。4.安装选项栏中，选择自定义，点击下一步。在命令提示符下，输入以下命令。下一步之后，就会进行安装。
【CUDA&cuDNN安装】深度学习基础环境搭建
qq_62223405的博客
06-21 2330
本文详细介绍了Windows系统下CUDA和cuDNN的安装配置流程。主要内容包括：1. 检查显卡型号与CUDA版本匹配性，解决驱动问题的方法；2. 分步骤指导CUDA 11.8的下载、安装和环境变量配置；3. 讲解cuDNN的下载安装和验证方法；4. 深入解析CUDA和cuDNN的功能区别：CUDA是通用GPU计算工具包，而cuDNN是专为深度学习优化的加速库。文章特别强调版本匹配的重要性，并提供了常见问题的解决方案，适合需要在Windows平台搭建深度学习环境的开发者参考。
CUDA、cuDNN安装
qq_36470994的博客
04-23 464
CUDA、cuDNN安装
Linux&anoconda安装tinycudann
qq_43697382的博客
03-26 672
首先要把这个repo clone过来一定别忘这个recursive如果你之前已经build失败过。需要先删除掉你编译过的东西。编译（其中第三行要等挺长时间的）链接pytorch检查是否安装成功。
cuda和cudnn的安装教程(全网最详细保姆级教程)
热门推荐
在职AI算法工程师，擅长计算机视觉，YOLO目标检测、分割等，擅长web、pyqt界面可视化，好内容持续更新中，来这里跟大家一起学习，共同进步
08-25 7万+
CUDA和cuDNN是由NVIDIA提供的两个关键软件库，用于利用NVIDIA GPU进行高性能计算和深度学习加速。
【配置】Ananconda 虚拟环境安装cuda、cudann
afdafassadada的博客
11-30 2502
python-annaconda
Windows下无需VS环境支持的Darknet GPU/CPU兼容版
这意味着用户不需要安装Visual Studio这个庞大的开发环境，也能够编译和运行darknet。虽然对于Windows环境下的C/C++项目来说，Visual Studio是常用的一个集成开发环境（IDE），但在某些情况下用户可能没有安装权限...
评论 1
成就一亿技术人!
拼手气红包6.0元
还能输入1000个字符
 
红包 添加红包
表情包 插入表情
表情包 代码片
 条评论被折叠 查看
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包

打赏作者

papaofdoudou

你的鼓励将是我创作的最大动力

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付：¥1
获取中
扫码支付

您的余额不足，请更换扫码支付或充值

打赏作者

实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值