惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Spread Privacy
Spread Privacy
K
Kaspersky official blog
Application and Cybersecurity Blog
Application and Cybersecurity Blog
Forbes - Security
Forbes - Security
Hacker News - Newest:
Hacker News - Newest: "LLM"
The Last Watchdog
The Last Watchdog
SecWiki News
SecWiki News
Attack and Defense Labs
Attack and Defense Labs
Google DeepMind News
Google DeepMind News
Security Archives - TechRepublic
Security Archives - TechRepublic
S
Secure Thoughts
WordPress大学
WordPress大学
Microsoft Security Blog
Microsoft Security Blog
P
Proofpoint News Feed
云风的 BLOG
云风的 BLOG
V
Visual Studio Blog
Security Latest
Security Latest
TaoSecurity Blog
TaoSecurity Blog
Cyberwarzone
Cyberwarzone
S
SegmentFault 最新的问题
Cloudbric
Cloudbric
aimingoo的专栏
aimingoo的专栏
S
Schneier on Security
N
Netflix TechBlog - Medium
MyScale Blog
MyScale Blog
T
The Blog of Author Tim Ferriss
H
Hacker News: Front Page
C
Cybersecurity and Infrastructure Security Agency CISA
小众软件
小众软件
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
AWS News Blog
AWS News Blog
AI
AI
G
GRAHAM CLULEY
IT之家
IT之家
P
Privacy & Cybersecurity Law Blog
L
Lohrmann on Cybersecurity
Last Week in AI
Last Week in AI
D
Docker
Recent Announcements
Recent Announcements
O
OpenAI News
T
Threat Research - Cisco Blogs
GbyAI
GbyAI
S
Security @ Cisco Blogs
T
Troy Hunt's Blog
C
Check Point Blog
博
博客园 - 三生石上(FineUI控件)
A
About on SuperTechFans
The Cloudflare Blog
阮一峰的网络日志
阮一峰的网络日志
N
News and Events Feed by Topic

kekxv 技术日志

基于 kekxv/gitea-pages 与 Gitea Actions 构建静态站点托管服务 Json简单工具 在Windows上运行Code Server：结合WSL打造你的云端VS Code开发环境 安卓sdkmanager工具换源 boost bazel starter bazel 供应商模式 PVE引导丢失修复 NSFW图像检测 警惕c++内置变量指针 关于内网springboot启动慢记录 网页转换为chrome插件 nginx代理的一种使用方式 YOLOv8 训练自己的数据 luckfox-交叉编译之bazel gitea actions CICD 自动化 Linux限制进程使用率 影音中心Jellyfin快速部署 OCR & 人脸算法 -- opencv dnn 深度学习记录-简单
tensorflow gpu 安装(ubuntu22.04)
kekxv · 2022-07-26 · via kekxv 技术日志

ubuntu22.04 安装 tensorflow-gpu 记录，主要是 NVIDIA 驱动麻烦，这里做个记录。

准备环境

先安装好好 python3 以及 python3-pip:

1
apt install python3 python3-pip

根据自己网络情况，是否使用清华源：

设为默认

升级 pip 到最新的版本 (>=10.0.0) 后进行配置：

1
2
pip install --upgrade pip
pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple

如果您到 pip 默认源的网络连接较差，临时使用本镜像站来升级 pip：

1
python -m pip install -i https://pypi.tuna.tsinghua.edu.cn/simple --upgrade pip

参考自： PyPI 镜像使用帮助 https://mirrors.tuna.tsinghua.edu.cn/help/pypi/

安装系统驱动

参考 ： 如何在Ubuntu 22.04 LTS上安装NVIDIA驱动程序

查看设备

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
$ ubuntu-drivers devices
== /sys/devices/pci0000:00/0000:00:02.0/0000:03:00.0 ==
modalias : pci:v0000000000000012313sv00000000000000123130000000000
vendor   : NVIDIA Corporation
model    : GM107 [GeForce GTX 750 Ti]
driver   : nvidia-driver-515-server - distro non-free
driver   : nvidia-driver-470 - distro non-free
driver   : nvidia-driver-390 - distro non-free
driver   : nvidia-driver-450-server - distro non-free
driver   : nvidia-driver-515 - distro non-free recommended
driver   : nvidia-driver-510-server - distro non-free
driver   : nvidia-driver-510 - distro non-free
driver   : nvidia-driver-418-server - distro non-free
driver   : nvidia-driver-470-server - distro non-free
driver   : xserver-xorg-video-nouveau - distro free builtin

安装

如果您对推荐版本感到满意，请使用以下命令：

1
ubuntu-drivers autoinstall

如果想安装指定版本(例如:nvidia-driver-515):

1
sudo apt install nvidia-driver-515

装好之后，重启系统，则可以加载驱动，使用显卡

目前（2022年07月26日） 2204 的库不全，需要使用 2004 的库

export last_public_key=3bf863cc.pub

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin

sudo mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600

sudo apt-key adv –fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/${last_public_key}

sudo add-apt-repository “deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/ /“

sudo apt-get update

sudo apt-get install libcudnn8

sudo apt-get install libcudnn8-dev

附2204版本：

1
2
3
4
5
6
7
8
wget https://developer.download.nvidia.cn/compute/cuda/repos/ubuntu2204/x86_64/cuda-ubuntu2204.pin
sudo mv cuda-ubuntu2204.pin /etc/apt/preferences.d/cuda-repository-pin-600
sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/3bf863cc.pub
sudo add-apt-repository "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/ /"
sudo apt-get update
sudo apt-get install cuda
sudo apt-get install libcudnn8
sudo apt-get install libcudnn8-dev

参考：https://stackoverflow.com/questions/66977227/could-not-load-dynamic-library-libcudnn-so-8-when-running-tensorflow-on-ubun

参考 ： https://www.tensorflow.org/install/pip

1
pip install --upgrade tensorflow-gpu

最新的pip仓库里面，可能会将tensorflow-gpu包删除，可以直接安装tensorflow，但是需要注意是否使用了GPU：

1
pip install --upgrade tensorflow

测试：

1
python -c "import tensorflow as tf;print(tf.reduce_sum(tf.random.normal([1000, 1000])))"

如果输出Could not load dynamic library 'libcudart.so.*'; 表示安装 cuda 失败了!!!

这时候可以考虑安装11的cuda版本：

1
apt install cuda-11-8

然后再试一次

成功的输出：

1
2
3
4
5
6
7
8
9
10
11
12
root@desktop:~# python3 -c "import tensorflow as tf;print(tf.reduce_sum(tf.random.normal([1000, 1000])))"
2022-07-26 13:49:48.585551: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-26 13:49:48.748572: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-26 13:49:48.748921: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-26 13:49:48.754087: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-26 13:49:48.754395: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-26 13:49:48.754683: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-26 13:49:49.974459: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-26 13:49:49.974697: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-26 13:49:49.974890: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-26 13:49:49.975066: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1532] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 1441 MB memory:  -> device: 0, name: NVIDIA GeForce GTX 750 Ti, pci bus id: 0000:03:00.0, compute capability: 5.0
tf.Tensor(2178.302, shape=(), dtype=float32)

此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。