Documentation
Rate This Document
Findability
Accuracy
Completeness
Readability
SubmitClose

Compiling PyTorch

  1. Obtain the PyTorch source code.
    1
    		 
    git clone -b v2.4.0 https://github.com/pytorch/pytorch.git --depth 1

    If a pull error occurs, try again or run the following command:

    1
    		 
    git config --global http.sslBackend openssl
  2. Access the PyTorch source code.
    1
    		 
    cd pytorch
  3. Obtain the sub-repository dependency code.
    1
2
    		 
    git submodule sync
git submodule update --init --recursive
  4. Set environment variables.
    1
2
3
4
    		 
    export USE_CUDA=1 
export USE_ROCM=1 
export USE_FBGEMM=0 
export USE_GLOO=1
  5. Install the Python dependency.
    1
    		 
    pip3 install -r requirements.txt
  6. Compile and install PyTorch.
    1
    		 
    python3 setup.py bdist_wheel

    After the installation is complete, the torch*.whl file is generated in the dist directory.

  7. Verify the installation.
    1
2
    		 
    pip3 install dist/torch*.whl
pip list | grep torch

    Command output:

  8. Check whether GPUs are supported.
    1
    		 
    python3

    Run the following commands:

    1
2
    		 
    import torch
print(torch.cuda.is_available())

    If the output is True, GPUs are supported.