Artificial intelligence configuration environment tutorial 2: Create a virtual environment in Anaconda Install the GPU version of Pytorch and torchvision and use the virtual environment in VsCode
- about the author
- 1. Check the CUDA version of your computer
- 2. Download and install CUDA
- 3. View environment variables
- 4. Create a virtual environment
- 5. Install the GPU version of Pytorch and torchvision
- 6. Using a virtual environment in VsCode
- Seven. Install the CPU version of Pytorch and torchvision
about the author
Meng Liping, female, School of Electronic Information, Xi'an Polytechnic University, 2021 graduate student, Zhang Hongwei's artificial intelligence research group.
Research direction: machine vision and artificial intelligence.
Email: [email protected]
This tutorial provides the installation packages of CUDA11.3, Pytorch1.10.0, and torchvision0.11.0 that need to be installed, and you can pick them up from the Baidu network disk link below!
Link: https://pan.baidu.com/s/18mgO8GtW1UnG6uijnnQvIQ?pwd=0843 Extraction code: 0843
– Shared from Baidu Netdisk super member V5
1. Check the CUDA version of your computer
According to the following two methods, check whether your computer supports CUDA, if not, then skip directly to [7]!
1.1 Method 1
1.1 .1 Use the shortcut key 'win + R'- - -> input cmd - - -> click OK
1.1.2 Enter the command 'nvidia-smi' in the terminal - - -> Check the CUDA version supported by your computer
The current CUDA version is 11.6, indicating that the highest supported version of CUDA is 11.6, which is backward compatible
1.2 Method 2
1.2.1 Right click on the desktop - - -> Open NVIDIA Control Panel
1.2.2 With help - - -> select system information
1.2.3 In Components - - -> View CUAN version
2. Download and install CUDA
NVIDIA official website: https://developer.nvidia.com/cuda-toolkit-archive
- Find the CUDA version suitable for your computer on the NVIDIA official website. It is recommended to install CUDA11.3 if the requirements are met; this tutorial provides the installation package of CUDA11.3
- Select the corresponding version of CUDA and download and install it;
- After downloading, install it in the default path;
CUDA installation reference post: https://blog.csdn.net/m0_45447650/article/details/123704930/ .
3. View environment variables
1. In Settings –> Search Advanced System Settings –> View Environment Variables
- The following four environment variables should appear in the environment system:
(This environment variable is for CUDA11.3 version, other versions need to correspond to the CUDA version you downloaded)
CUDA_PATH
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.3
CUDA_PATH_V11_3
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.3
NVCUDASAMPLES_ROOT
C:\ProgramData\NVIDIA Corporation\CUDA Samples\v11.3
NVCUDASAMPLES11_3_RO…
C:\ProgramData\NVIDIA Corporation\CUDA Samples\v11.3
- If there is no corresponding environment variable, you need to add it manually
Verify that CUDA is successfully installed
-
Run cmd and enter 'nvcc -v' to view the version number;
-
Enter 'set cuda' to view the environment variables set by CUDA. The
following figure shows that CUDA is successfully installed. CUDA
installation reference post: https://blog.csdn.net/m0_45447650/article/details/123704930/ .
4. Create a virtual environment
4.1 Create a virtual environment using instructions
- Add mirror source:
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/win-64/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/win-64/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/win-64/
conda config --set show_channel_urls yes
conda config --set ssl_verify false
- Create a virtual environment, install Python 3.6 version in this tutorial; run the following commands on the local cmd terminal:
conda create -n py36 python=3.6
4.2 Check whether the created virtual environment appears in Anconda
Open Anconda, click Environments to see if there is a virtual environment named py36
Note: Next, install various scientific computing packages and frameworks (Pytorch) in a virtual environment named py36. If there are multiple environments, for example, python3.7 is required, and a virtual environment of python3.7 is created. Conda management will not There is a version conflict.
4.3 Virtual environment related instructions
a : The terminal runs the following command: View the local environment
conda env list
b: The terminal runs the following command: enter the virtual environment
conda activate py36
c: The terminal runs the following command: Exit the virtual environment
conda deactivate
Precautions: Enter the virtual environment as shown in the figure: there will be parentheses (virtual environment name: py36), after entering the virtual environment, you can configure Pytorch in an environment where Python is 3.6, and install it as much as you like!
5. Install the GPU version of Pytorch and torchvision
This tutorial installs CUDA11.3, Pytorch1.10.0, torchvision0.11.0
There are two ways to install Pytorch and torchvision, online and offline; if there is a problem that the installation fails in the online way, you can choose to install offline, and the corresponding installation package will be provided.
5.1 Online installation
- Install the mirror source: execute the following five instructions in sequence;
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/win-64/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/win-64/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/win-64/
conda config --set show_channel_urls yes
conda config --set ssl_verify false
-
Enter the created virtual environment (py36) in the local terminal, when (py36) appears at the top, it indicates that the virtual environment of py36 has been entered;
- Enter the official website of Pytorch: https://pytorch.org/ -
Click on previous Pytorch versions
-
Find the Pytorch installation instruction under CUDA11.3 under Windows system and copy it;
If the installed CUDA version is not 11.3, you need to select the corresponding version of Pytorch and torchvision
-
Paste in the py36 virtual environment of the cmd local terminal to run the installation:
conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=11.3 -c pytorch -c conda-forge
5.2 Offline installation
5.2.1 Download and install torch
Pytorch download link: https://download.pytorch.org/whl/torch_stable.html .
-
Select the corresponding version of torch and download it
-
Enter the following command in the local terminal to install offline. When Successfully installed... appears , it indicates that the installation is successful; use any one of the following three commands;
pip install F:\Demo\torch\torch-1.10.0+cu113-cp36-cp36m-win_amd64.whl
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple F:\Demo\torch\torch-1.10.0+cu113-cp36-cp36m-win_amd64.whl
pip install F:\Demo\torch\torch-1.10.0+cu113-cp36-cp36m-win_amd64.whl http://pypi.douban.com/simple/trusted host pypi.douban.com
Remarks: "F:\Demo\torch\torch-1.10.0+cu113-cp36-cp36m-win_amd64.whl" in the above command indicates the path + file name to download the torch installation package and save it on your computer; download according to your own Changes to saved conditions;
5.2.2 Download and install torchvision
Pytorch, torchvision and python have version correspondence
- The version correspondence between Pytoch and torchvision can be queried on the official website of Pytorch , as follows:
This tutorial found that the torchvision version corresponding to Pytoch1.10.0 is 0.11.0 .
- Offline download torchvision 0.11.0
torchvision download link: https://download.pytorch.org/whl/torch_stable.html .
- Use the pip command to install torchvision0.11.0; use any of the following three commands;
pip install F:\Demo\torchvision\torchvision-0.11.0+cu113-cp36-cp36m-win_amd64.whl
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple F:\Demo\torchvision\torchvision-0.11.0+cu113-cp36-cp36m-win_amd64.whl
pip install F:\Demo\torchvision\torchvision-0.11.0+cu113-cp36-cp36m-win_amd64.whl http://pypi.douban.com/simple/trusted host pypi.douban.com
Note: "F:\Demo\torchvision\torchvision-0.11.0+cu113-cp36-cp36m-win_amd64.whl" in the above command indicates the path + file name of the torchvision installation package; change it according to your own situation;
- If torch is automatically updated during the installation of torchvision, use the following command to install torchvision;
pip install --no-deps F:\Demo\torchvision\torchvision-0.11.0+cu113-cp36-cp36m-win_amd64.whl
5.3 Verify successful installation
- After the installation is complete, run the following commands in sequence in the virtual environment to verify whether Pytoch is successfully installed and whether the GPU is available;
python
import torch
torch.cuda.is_available()
If import torch does not report an error, it means that torch is installed successfully
If the result of torch.cuda.is_available() is True, it means that the GPU version is installed
6. Using a virtual environment in VsCode
6.1 Install the Python plugin
- Enter the VsCode software, click "Extension", enter Python in the search box, and then select Install in the lower right corner of the Python plug-in;
6.2 Add a virtual environment
-
Press the shortcut key "Ctrl+Shift+P" to bring up the global setting search window, and then enter "Python:Select Interpreter" and the "Python:Select Interpreter" option will appear, click this option;
-
Click this option to jump to the Python interpreter configuration window, where the added Python interpreter is displayed, select py36, and you can use Python in VsCode;
-
Note: If VsCode enters the terminal, it is a PS environment; Then first output cmd in the terminal , exit the PS environment, and then manually enter "conda activate py36" to enter the configured py36 environment; you can view the installed torch and torchvision by entering "pip list", and confirm that you downloaded and installed it yourself version of
The above is all about creating a virtual environment in Anaconda, installing the GPU version of Pytorch and torchvision, and using the virtual environment in VsCode! ! !
Next is the installation tutorial of the CPU version of Pytorch! ! !
Prerequisite: Complete step 4 first. Create a virtual environment
Seven. Install the CPU version of Pytorch and torchvision
- Enter the virtual environment, use the Pytoch official website to install and run according to the instructions;
conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cpuonly -c pytorch
- Verify that the installation is successful, and import torch succeeds without reporting an error:
python
import torch
The above is all about creating a virtual environment in Anaconda, installing the CPU version of Pytorch and torchvision! ! !