GPU server fool-proof installation of Anaconda, pytorch, tensorflow

1. Download Anaconda installation package, official website address, Tsinghua source address.

After downloading from the official website to the local, you can upload the installation package to the server through file transfer, and use the Tsinghua source address to directly use wget to download the required version, for example:

wget https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/Anaconda3-2022.05-Linux-x86_64.sh

Tsinghua University Open Source Software Mirror Station
Anaconda official website download
wget download Anacondalinux installation package

2. Install Anaconda

bash Anaconda3-2022.05-Linux-x86_64.sh

Then, Enter
Install Anaconda
Accept the license terms, yes, then information about the installation location of Anaconda3 will be displayed.
Installation returns information
Return information during installation
Return information during installation
Return information during installation
Follow the prompts to activate the basic environment of conda

eval "$(/home/cxcai/anaconda3/bin/conda shell.bash hook)"
conda init

conda init

conda info # View conda related information

conda info

3.conda change source

conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
conda config --set show_channel_urls yes # Display channel address when setting search

conda config –add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
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/r/
conda config –add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/
conda config –add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/msys2/
conda config –add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda/
conda config –add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/menpo/
conda config –add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/

conda change source

Note: restore the original channel command

conda config --remove-key channels # restore channel

4. Install pytorch

  • Query cuda version
nvidia-smi

  • Install the corresponding pytorch version, check the pytorch version corresponding to cuda on the official website, or use conda search pytorch to query the pytorch version that can be installed (downward compatible)


  • conda creates and activates a virtual environment
conda create -n envpytorch python==3.10 # Create envpytorch virtual environment and configure python3.10
source activate envpytorch # activate the virtual environment
conda activate envpytorch # enter the virtual environment
conda deactivate # Exit the virtual environment
  • Install pytorch, torchvision, torchaudio, cudatoolkit in the created virtual environment
  • Check if the installation is successful
  • Install matplotlib, pandas, numba, seaborn and other libraries
conda install matplotlib
conda install pandas
conda install numba
conda install seaborn

Install TensorFlow

  • Also create and activate the virtual environment trf
conda create -n trf python==3.8.16
source activate trf
conda activate trf
conda deactivate trf
conda env list # View all virtual environments

  • Check the corresponding tensorflow-gpu version, or use conda search tensorflow-gpu

  • Install tensorflow-gpu2.4.1
  • After the installation is complete, test whether tensorflow is installed successfully
python
import tensorflow as tf

print(tf.__version__)
print(tf.test.gpu_device_name())
print(tf.config.experimental.set_visible_devices)
print('GPU:',tf.config.list_physical_devices('GPU'))
print('CPU:',tf.config.list_physical_devices(device_type='CPU'))
print(tf.config.list_physical_devices('GPU'))
print(tf.test.is_gpu_available())
# output the number of GPUs available
print("Num GPUs Available: ", len(tf.config.experimental.list_physical_devices('GPU')))

It can be seen that there is an error in this installation


The reason for the error is that numpy does not have an object attribute. After consulting the materials, it is found that the numpy version is too high, so choose to reduce the version

conda install numpy==1.23.4



Test again whether the installation was successful

The installation is successful! ! ! ! ! ! ! ! ! ! ! ! ! !