Install TensorFlow-gpu 2.0 on Anaconda for Windows 10/Ubuntu

Update:

Update #5 (23/02/2023): This post is an old post now, I haven’t updated it since I moved to Pytorch.
Please follow the guide Install TensorFlow with pip on the TensorFlow website.

Update #4 (01/11/2019): Anaconda is supporting the Tensorflow 2.0.0. Then everything is simple now. Thank you @Vorphus

Update #3 (01/10/2019): TensorFlow 2.0 has been released! But Anaconda still hasn’t supported yet so I think we will have to install by pip until Anaconda officially supports.

Update #2 (18/08/2019): The latest version Anaconda (Anaconda2019.07) uninstall all the tensorflow relative packages (including cudnn, cupti and cudatoolkit). So I will remove the redundant step “installing tensorflow 1.14”.

I think installing cudnn, cupti and cudatoolkit are enough.

Update #1 (08/06/2019): TensorFlow 2.0 Beta is available and Anaconda has cudnn 7.6.0 so we don’t have to manually copy cudnn to our environment anymore.

This is a tutorial to install TensorFlow-gpu 2.0 since Anaconda has supported Tensorflow 2.0.0.

1/ Download and install Anaconda for Window or Ubuntu:
https://www.anaconda.com/distribution/

2/ After installed Anaconda, open the Anaconda Prompt.

Now, Anaconda is supporting Tensorflow 2.0.0 so we can create the environment containing TF2 directly.

conda create -n your_env_name python=3.6 tensorflow-gpu

3/ Activate your_environment

conda activate your_env_name

Then it’s ok. You can run the sample projects of TensorFlow-gpu 2.0 on Anaconda.