CUDA, cuDNN and Tensorflow

Installing cuDNN on Linux

1. Installing NVIDIA Graphics Driver

2. Installing CUDA

   2.1 Install repository meta-data -
      sudo dpkg -i cuda-repo-<distro>_<version>_<architecture>.deb

   2.2 Install CUDA public GPG-key

             sudo apt-key add /var/cuda-repo-<version>/7fa2af80.pub

   2.3 Update apt repository cache
             sudo apt-get update

   2.4 Install CUDA
             sudo apt-get install cuda

   2.5 Environment setup
             export PATH=/usr/local/cuda-10.0/bin${PATH:+:${PATH}}

3  Downloading cuDNN

   3.1 Navigate to your <cudnnpath> directory containing cuDNN Debian file.

   3.2 Install the runtime library, for example:
       sudo dpkg -i libcudnn7_7.0.3.11-1+cuda9.0_amd64.deb

   3.3 Install the developer library, for example:
       sudo dpkg -i libcudnn7-dev_7.0.3.11-1+cuda9.0_amd64.deb

   3.4 Install the code samples and the cuDNN Library User Guide, for example:
       sudo dpkg -i libcudnn7-doc_7.0.3.11-1+cuda9.0_amd64.deb

Check the version of CUDA
cat /usr/local/cuda/version.txt

Check the version of cuDNN
cat /usr/include/cudnn.h | grep CUDNN_MAJOR -A 2