In the past, the environment of the laboratory server was installed directly through pip, and anaconda was not used.
When using the server in the laboratory these days, I found that the environment I built was modified by the same door. . .
My whole person was split. I learned from the pain and decided to secretly install an anaconda and create a virtual environment to avoid being modified by the same door again.
I won’t go into details about how to set up ssh. Just start installing anaconda. Refer to ssh to connect to the Linux server and install Anaconda.
Visit the official website of anaconda under windows
Right click 64-Bit (x86) Installer (529 MB) , click copy link address to get the download link
In the linux command line, download through wget
wget https://repo.anaconda.com/archive/Anaconda3-2020.11-Linux-x86_64.sh
If prompted that the permissions are not enough, add sudo
sudo wget https://repo.anaconda.com/archive/Anaconda3-2020.11-Linux-x86_64.sh
After the download is complete, you can find the installation package in the directory:
by
ll
With the Anaconda installation package, then install Anaconda
Since the uploaded file does not have executable permissions, you need to add executable permissions to the file first.
chmod u+x Anaconda3-2020.11-Linux-x86_64.sh
If prompted that the permissions are not enough, also add sudo
sudo chmod u+x Anaconda3-2020.11-Linux-x86_64.sh
With executable permissions, directly execute the "Anaconda3-2020.11-Linux-x86_64" file
./Anaconda3-2020.11-Linux-x86_64.sh
The following interface is displayed
Press Enter to continue
Read the software license,
Just press "q" directly,
Then enter "yes" to accept the license.
Specify the installation path of Anaconda.
Generally established under the user of home, mine is
/home/zhanglei/conda
Note that the implementation does not need to create the conda folder in advance and write the address. If you make a mistake, a strange symbol will appear when you press the backspace key. You need to hold down ctrl and then press backspace to delete
After typing the address, press enter to confirm
After installation
Select yes in whether to add Anaconda environment variables
Reconnect ssh and input "conda -V" to successfully output the version of conda
After Anaconda is installed, create a virtual environment
by
conda create -n xxx python=3.6
Where xxx is your environment name
For example, mine is
conda create -n tensorflow-gpu python=3.6
If you need to delete the environment
use
conda remove -n tensorflow-gpu --all
tensorflow-gpu is my environment name
After setting up the environment, activate the environment
conda activate your_env_name(虚拟环境名称)
或者
source activate your_env_name(虚拟环境名称)
After activating the environment, we can install tensorflow-gpu
If you need to increase the domestic mirror source
Need to use
conda config --add channels
Instruction
Such as adding Tsinghua sources
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
Similarly, you can increase the source of Jiaotong University
conda config --add channels https://mirrors.sjtug.sjtu.edu.cn/anaconda/pkgs/main/
conda config --add channels https://mirrors.sjtug.sjtu.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.sjtug.sjtu.edu.cn/anaconda/cloud/conda-forge/
University of Science and Technology
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/pkgs/main/
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge/
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/cloud/msys2/
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/cloud/bioconda/
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/cloud/menpo/
Remember to add
conda config --set show_channel_urls yes
It means to display the url of the channel when installing the package from the channel, so that you can know the source of the package installation.
You can use Anaconda to install tensorflow-gpu directly under Anaconda without installing cuda cudnn separately
conda install tensorflow-gpu
If you need to limit the version of tensorflow (such as 1.15), you can change the code to
conda install tensorflow-gpu==1.15
Anaconda will automatically download cuda+cudnn
Start after typing y
This shows that the installation was successful
If there is no automatic installation, for example, I found that tensorflow-gpu 1.14 will not automatically install cuda and cudnn, so I need to install it manually.
You can first see which versions are in the mirror source
conda search cudnn
conda search cudatoolkit
Download according to the version number
conda install cudatoolkit=10.0.0
conda install cudnn=7.4
The version number needs to be selected according to the compatibility relationship
You can also pass
conda list
View installed libraries
You can see that tensorflow-gpu 1.15 and the supporting cuda10 have been installed
Let's check if it can be used normally
enter
python3
import tensorflow as tf
tf.test.is_gpu_available()
Can return true to indicate successful installation
If you need to exit your virtual environment, please enter
conda deactivate
If you need to view the specific location of the environment, please enter
conda env list
I have the basic base environment and the tensorflow-gpu environment just set
Pay attention to the address of this environment, you need to use pycharm to remotely call the virtual environment under the server Anaconda
Next, you can connect to the server with pycharm
Click on tools-->Deployment-->Configuration
Create a new SFTP connection
And set the mapping
Then enter File-->settings
In the python interpreter under project:xxx (your project name), click the gear, and then click show all
Click the plus sign
Select the set ssh configuration under ssh interpreter
Then choose Sync folders according to your preference, I usually save it under the mnt folder
Note that the Interpreter here should select python3 under the address displayed by the conda env list before
After selecting, click Finish
Then, we can try it in pycharm, can the remote call succeed
enter
import tensorflow as tf
print(tf.test.is_gpu_available())
We see the result
The graphics card is read and true is returned
Description can be successfully called
You will find that after installing anaconda, as long as you enter ssh, there will be (base) at the beginning of each line. . . This of course is not in line with my purpose of secretly pretending to be anaconda
You can hide (base) by setting the auto_activate_base parameter to false:
conda config --set auto_activate_base false
2. Pass if you want to enter
conda activate base
3. If you regret it or want to keep the base, then pass
conda config --set auto_activate_base true
To recover
Referring After installing anaconda, the system displays the front of the system (base) word