Install cuDNN v8.0.5 in ubuntu 20.04

• what: NVIDIA cuDNN是用于深度神经网络的GPU加速库。cuDNN安装很简单，就是将解压的文件放到指定位置即可.
• why: GPU acceleration for DNN
• how: cuDNN安装很简单，就是将解压的文件放到指定位置即可[2].
  • step 1: Register a Nvidia account and go to the download page. Select the version based on your CUDA version
  • step 2: navigate to your directory containing the tar file, and unzip the cuDNN via tar -xzvf cudnn-x.x-linux-x64-v8.x.x.x.tgz
  • step 3: copy files to target location

    $ sudo cp cuda/include/cudnn*.h /usr/local/cuda/include
    $ sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64
    $ sudo chmod a+r /usr/local/cuda/include/cudnn*.h /usr/local/cuda/lib64/libcudnn*
    

  • step 4: check result via cat /usr/local/cuda/include/cudnn_version.h | grep CUDNN_MAJOR -A 2
  • step 5: Then downloaded the other 3 .deb files and install them:

    $ cd ~/Downloads   
    $ sudo dpkg -i .deb
    $ sudo dpkg -i .deb
    $ sudo dpkg -i .deb
    

  • step 6: verifying the install on Linux [2]. If cuDNN is properly installed and running on your Linux system, you will see a message similar to the following:
    Test passed!

    # Copy the cuDNN samples to a writable path.
    $ cp -r /usr/src/cudnn_samples_v8/ $HOME
    # Go to the writable path.
    $ cd  $HOME/cudnn_samples_v8/mnistCUDNN
    # Compile the mnistCUDNN sample.
    $ make clean && make
    # Run the mnistCUDNN sample.
    $ ./mnistCUDNN
    # If cuDNN is properly installed and running on your Linux system, you will see a message similar to the following:
    # Test passed!
    

• record:
  • step 1: download the latest version Download cuDNN v8.0.5 (November 9th, 2020), for CUDA 10.1. (Q: will there be problemetic when using PyTorch built in cudnn with v7.6.5?)
  • step 2: unzip

    $ cd ~/Downloads
    $ tar -xzvf cudnn-10.1-linux-x64-v8.0.5.39.tgz
    

  • step 3: done, and every command is okay
  • step 4: done, and got 8.0.5
  • step 5: done.
  • step 6: got Test passed!
• result: pass
• ref:
  • [1] Ubuntu:安装cudnn10.1
  • [2] Official instllation guide of cudnn