What dependencies are required for CentOS to install PyTorch

Deploying the PyTorch deep learning framework on a CentOS system requires several dependencies to be installed in advance. To ensure the best development experience, it is recommended to use Anaconda3 as an environment management tool.

1. Dependency installation:

1. Anaconda3: Download and install the Anaconda3 version compatible with the CentOS system from the official Anaconda website.
2. glibc: Ensure that the system glibc version is not lower than 2.17. You can use rpm -qa | grep glibc command to view the current version. If the version is too low, you need to update the system or install a higher version of glibc.
3. Development tools: Install necessary compilation tools such as gcc and make. You can usually install it using the yum command: sudo yum groupinstall "Development Tools"
4. Python: It is recommended to use Python versions 3.6 to 3.9. After the installation of Anaconda3 is completed, it will bring its own Python environment.
5. pip: Anaconda3 comes with its own pip package manager, no additional installation is required.
6. CUDA and cuDNN (optional): If you need to accelerate PyTorch operations with GPU, you must install CUDA and cuDNN compatible with PyTorch. Please visit NVIDIA's official website, download the corresponding version of the CUDA Toolkit and cuDNN library, and install it according to the official documentation.

2. PyTorch installation steps:

1. Create an Anaconda virtual environment: Use the conda command to create a virtual environment named pytorch and specify the Python version (for example, 3.8):

   conda create -n pytorch python=3.8
   conda activated pytorch

2. Install PyTorch: In the activated pytorch environment, use conda to install PyTorch and its related libraries (torchvision, torchaudio). If GPU support is required, please select the appropriate cudatoolkit version number according to your CUDA version. For example, if your CUDA version is 11.3:

   conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch

   Note: The version number of cudatoolkit must match the installed CUDA version. You can use the conda info cudatoolkit command to view available CUDA versions. If there is no GPU support, cudatoolkit parameter is ignored.

3. Verify the installation: After the installation is complete, run the following Python code to verify that PyTorch is installed successfully and whether the GPU is available:

   import torch
   print(torch.__version__)
   print(torch.cuda.is_available())

   If the output displays the PyTorch version number and torch.cuda.is_available() returns True (in the case of GPU), the installation is successful.

Through the above steps, you can successfully install and run PyTorch on your CentOS system. Please make sure that your system configuration meets the requirements of PyTorch, and refer to the official PyTorch documentation for the latest installation guide and compatibility information.

The above is the detailed content of What dependencies are required for CentOS to install PyTorch. For more information, please follow other related articles on the PHP Chinese website!