Download Version 3.6 64bit here: Windows x86-64 executable installer

 cd C:\Users\%USERNAME%\AppData\Local\Programs\Python\Python36\Scripts

if you tensorflow already installed so uninstalled :

 pip uninstall tensorflow

 pip install tensorflow-gpu

Download Last Version here: Nvidia Last Driver

Download Version 9 here: CUDA Toolkit 9.0

1. Download Version 7.0.5 here: cuDNN v7.0.5 (Dec 5, 2017), for CUDA 9.0

2. unzip cudnn-9.0-windows10-x64-v7.zip then copy bin &include & lib in :

 C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0

1. Click + R Then type sysdm.cpl and go to advanced and select Environmental Variables

2. Click on the Path in system variable and select edit

3. add the following paths

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0\bin
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0\libnvvp

GPU Testing

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

 [name: "/device:CPU:0" device_type: "CPU" memory_limit: 268435456 locality { } incarnation: 7179507163498351914 , name: "/device:GPU:0" device_type: "GPU" memory_limit: 2241268940 locality { bus_id: 1 } incarnation: 14675117801653382741 physical_device_desc: "device: 0, name: GeForce GTX 1060 3GB, pci bus id: 0000:23:00.0, compute capability: 6.1" ]