Tensorflow GPU Installation on Ubuntu 18.04

This article is about complete installation step for Tensorflow-GPU on Ubuntu 18.04 .As we can check that NVIDIA have supported driver and CUDA version for respective NVIDIA product.

Step1: NVIDIA driver version

First check what is the version of NVIDIA driver on your GPU system.You can check it with below command.

nvidia-smi

snippet after running nvidia-smi on my system.

As like same you have to cross check driver version and if it is not there ,you have to install first.

Steps for installation of NVIDIA driver:

Click here to select your NVIDIA product information to select graphics driver and follow the instructions.

In my case , I have given above information.

Step2: Install CUDA toolkit

First check here recommended version for tensorflow, cuDNN and CUDA .

image represent recommended version

For example if you are using Tensorflow version 1.12.0 then CUDA version 9.0 and cuDNN 7.0 is recommended.

selected version for Linux-Ubuntu18.04

selected version

Once you downloaded the file, you have to run below command.

sudo dpkg -i cuda-repo-ubuntu1804–10–0-local-10.0.130–410.48_1.0–1_amd64.deb

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

sudo apt-get update

sudo apt-get install cuda

In the installation step , make sure that again you do not agree to install new driver.

Step3 : Install CUDNN

Click here and select the option ‘Download cuDNN’ then join or login with your membership details and find the recommended version of cuDNN as given above.

select archive file

Once you get download your CUDA version archive , run below command.

# Unpack the archive
tar -zxvf cudnn-9.0-linux-x64-v7.tgz
# Move the unpacked files to your CUDA directory
sudo cp -P cuda/lib64/libcudnn* /usr/local/cuda-9.0/lib64/
sudo cp  cuda/include/cudnn.h /usr/local/cuda-9.0/include/
# Change file access to all users
sudo chmod a+r /usr/local/cuda-9.0/include/cudnn.h /usr/local/cuda/lib64/libcudnn*

Step 4: Install libcupti

Run below command for this.

sudo apt-get install libcupti-dev

Step 5: Set the Environment Variable

On command prompt tensorflow will find the CUDA installation to use for GPU enabling. You have to set your environment variable for this.

Open .bashrc file and append below lines and source to the file.

export PATH=/usr/local/cuda-9.0/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

After source , restart your terminal.

Step 6: Install your Tensorflow-GPU

For Python3 version:

pip3 install –upgrade tensorflow-gpu

For Python 2 version :

pip install –upgrade tensorflow-gpu

Step7: Verify your installation

With the installed version of your python , run below command on ternimal and verify that your Tensorflow is running on the GPU.

Python3 
from tensorflow.python.client import device_lib
device_lib.list_local_devices()

snippet for above command run

Note: We should check the compatibility of Tensorflow ,CUDA and cuDNN in case of any particular application.

Happy learning…