日萌社
人工智能AI:Keras PyTorch MXNet TensorFlow PaddlePaddle 深度学习实战(不定时更新)
Ubuntu 18.04.4、CUDA、cuDNN、anaconda、pycharm、tensorflow、pytorch下载安装包
链接:https://pan.baidu.com/s/1oqYxOYob9MBuqHYsxkS-IA
提取码:y8ub
链接:https://pan.baidu.com/s/1YqfX0ObJSSUIaHYW3OW3cQ
提取码:21lu
链接:https://pan.baidu.com/s/1yF7e6ntWpXpdPWFLN4gi2g
提取码:f1zi
安装vim
sudo apt-get install vim
安装ifconfig
sudo apt install net-tools
ifconfig
更新软件源和升級软件
sudo apt-get update 更新软件源
sudo apt-get upgrade 升級软件
安装gcc
sudo apt install build-essential
#该命令将安装一堆新包,包括gcc,g++和make。
#要验证GCC编译器是否已成功安装,请使用gcc --version命令打印GCC版本:
gcc --version
JDK1.8 下载安装
1.下载:https://www.oracle.com/java/technologies/javase-jdk8-downloads.html
2.卸载自带的OpenJDK方法 sudo apt-get remove openjdk*
3.安装 JDK 1.8
sudo mkdir /usr/local/java
sudo tar zxvf jdk-8u251-linux-x64.tar.gz -C /usr/local/java
cd /usr/local/java/jdk1.8.0_251/bin
./java -version
4.vim ~/.bashrc
1.在文件末尾添加以下内容:
#set java environment
JAVA_HOME=/usr/local/java/jdk1.8.0_251
CLASSPATH=.:$JAVA_HOME/lib/tools.jar
PATH=$JAVA_HOME/bin:$PATH
export JAVA_HOME CLASSPATH PATH
2.重新加载profile配置文件,让配置文件生效:source ~/.bashrc
3.检查是否已配置好新的JDK:java -version
CUDA、tensorflow版本清单:https://tensorflow.google.cn/install/source#linux
CUDA 全部版本下载:https://developer.nvidia.com/cuda-toolkit-archive
cuda 10.1下载:https://developer.nvidia.com/cuda-10.1-download-archive-update2
迅雷下载cuda 10.1:http://developer.download.nvidia.com/compute/cuda/10.1/Prod/local_installers/cuda_10.1.243_418.87.00_linux.run
命令下载cuda 10.1
wget http://developer.download.nvidia.com/compute/cuda/10.1/Prod/local_installers/cuda_10.1.243_418.87.00_linux.run
安装:sudo sh cuda_10.1.243_418.87.00_linux.run
配置环境变量:
vim ~/.bashrc
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-10.1/lib64
export PATH=$PATH:/usr/local/cuda-10.1/bin
export CUDA_HOME=$CUDA_HOME:/usr/local/cuda-10.1
source ~/.bashrc 刷新
nvcc -V 查看版本
cuDNN下载(必须先注册账号登录才能下载):https://developer.nvidia.com/rdp/cudnn-download
Windows 10版本直接下载:https://developer.nvidia.com/compute/machine-learning/cudnn/secure/7.6.5.32/Production/10.1_20191031/cudnn-10.1-linux-x64-v7.6.5.32.tgz
Linux版本直接下载:https://developer.nvidia.com/compute/machine-learning/cudnn/secure/7.6.5.32/Production/10.1_20191031/cudnn-10.1-windows10-x64-v7.6.5.32.zip
安装:tar -zxvf cudnn-10.1-linux-x64-v7.6.5.32.tgz.tgz
sudo cp ./cuda/include/cudnn.h /usr/local/cuda/include
sudo cp ./cuda/lib64/libcudnn* /usr/local/cuda/lib64
sudo chmod 777 /usr/local/cuda/include/cudnn.h
sudo chmod 777 /usr/local/cuda/lib64/libcudnn*
anaconda python 3.7
1.安装:bash Anaconda3-2019.10-Linux-x86_64.sh
安装过程只需要按回车和yes即可
2.配置环境变量
打开.bashrc文件vim ~/.bashrc
文件末尾添加一行 export PATH="/home/用户名/anaconda3/bin:$PATH"
使配置马上生效 source ~/.bashrc
3.查看anaconda版本信息:conda --version
4.更新pip版本:python -m pip install --upgrade pip
5.查看pip版本:pip show pip
pycharm安装
tar zxvf pycharm-professional-2019.1.3.tar.gz
cd ./pycharm-2019.1.3/bin
sh ./pycharm.sh
验证:
1.linux添加:-javaagent:/home/nagisa/pycharm-2019.1.3/bin/jetbrains-agent.jar
window添加:-javaagent:G:\PyCharm 2019.2.3\bin\jetbrains-agent.jar
2.Window/linux中hosts无需添加0.0.0.0 account.jetbrains.com和0.0.0.0 www.jetbrains.com,否则可能报错误码1653219
Linux:vim /etc/hosts
Window:C:\Windows\System32\drivers\etc\hosts
把下面两行注释掉,无需添加到hosts中
0.0.0.0 account.jetbrains.com
0.0.0.0 www.jetbrains.com
重新加载网络 sudo /etc/init.d/networking restart
tensorflow gpu 2.1.0 安装
1.添加国内的下载源会提下载速度 -i https://pypi.tuna.tsinghua.edu.cn/simple
pip install tensorflow-gpu==2.1.0 -i https://pypi.tuna.tsinghua.edu.cn/simple
2.国内源直接下载链接
选择下载:tensorflow_gpu-2.1.0-cp37-cp37m-manylinux2010_x86_64.whl
下载链接:https://pypi.tuna.tsinghua.edu.cn/packages/0c/e8/56ecca076108302a0bc34d509dc891086455ac31895843403ef0a71d0497/tensorflow_gpu-2.1.0-cp37-cp37m-manylinux2010_x86_64.whl (421.8 MB)
安装:pip install tensorflow_gpu-2.1.0-cp37-cp37m-manylinux2010_x86_64.whl
pip list
tensorboard 2.1.1
tensorflow-estimator 2.1.0
tensorflow-gpu 2.1.0
3.测试gpu
import tensorflow as tf
print(tf.test.is_gpu_available()) #返回True
print(tf.__version__)
pytorch gpu 安装
1.官网下载安装:https://pytorch.org/get-started/locally/
安装命令:conda install pytorch torchvision cudatoolkit=10.1 -c pytorch
2.国内源直接下载链接
网站链接:
window:https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/win-64/
linux:https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/linux-64/
选择下载:
window:pytorch-1.5.0-py3.7_cpu_0.tar.bz2
linux:pytorch-1.5.0-py3.7_cuda10.1.243_cudnn7.6.3_0.tar.bz2
下载链接:
window:https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/win-64/pytorch-1.5.0-py3.7_cpu_0.tar.bz2
linux:https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/linux-64/pytorch-1.5.0-py3.7_cuda10.1.243_cudnn7.6.3_0.tar.bz2
安装:
window:conda install pytorch-1.5.0-py3.7_cpu_0.tar.bz2
linux:conda install pytorch-1.5.0-py3.7_cuda10.1.243_cudnn7.6.3_0.tar.bz2
pip list
torch 1.5.0
3.测试GPU版本的代码
import torch
#测试 CUDA
print("Support CUDA ?: ", torch.cuda.is_available()) #返回True
x = torch.tensor([10.0])
x = x.cuda()
print(x)
y = torch.randn(2, 3)
y = y.cuda()
print(y)
z = x + y
print(z)
# 测试 CUDNN
from torch.backends import cudnn
print("Support cudnn ?: ",cudnn.is_acceptable(x)) #返回True