Tensorflow Installation via Wheel file in a minute

1. Download the required tensorflow version (whl file) from the below link:-

https://pip.aws.lolatravel.com/pip/dev/+simple/tensorflow-gpu/


Please check the compatibility with versions on https://www.tensorflow.org/install/source#common_installation_problems

2. Create environment using conda

conda create -n tensorflow1_4 python=2.7/3.5

3. Activate the environment

source avtivate tensorflow1_4


4. Install using the following command

pip install --ignore-install tensorflow-gpu-1.4.1-cp35-cp35m-manylinux1_x86_64.whl



Code to check tensorflow-gpu is installed or not

import tensorflow as tf

tf.test.is_gpu_available(
    cuda_only=False,
    min_cuda_compute_capability=None
)

It will return True if installed 

otherwise, false

You might get this error:- 

ImportError: libcudnn.so.5/6/7: cannot open shared object file: No such file or directory

To Install CUDnn packages locally visit the following site:-

https://developer.nvidia.com/rdp/cudnn-archive

Update bashrc with LD_LIBRARY_PATH

export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64"