Tensoflow-gpu

https://www.codingforentrepreneurs.com/blog/install-tensorflow-gpu-windows-cuda-cudnn/

Install Anaconda3 3.52 with python 3.6.5

https://repo.anaconda.com/archive/Anaconda3-5.2.0-Windows-x86_64.exe


Install Visual Studio 2015 Community

Download Older Visual Studio Software | Visual Studio

CUDA v10.0
cudnn 7.4.1.5

Update your system environment variables’ PATH to have:

  • C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\bin
  • C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\libnvvp

python -m pip install –upgrade pip
pip install msgpack
pip install tensorflow-gpu==2.0.0-alpha0

Testcode: python
>>> from tensorflow.python.client import device_lib
>>> print(device_lib.list_local_devices())

And

https://gist.githubusercontent.com/mrry/ee5dbcfdd045fa48a27d56664411d41c/raw/e10575d2583a896a6fab08251ce74a73520b0a06/tensorflow_self_check.py

Tested Working:Win10 +VS2015+ python3.65 + CUDA10.0 + cuDNNv7.4.1 + tensorflow-gpu 1.12.0

Bechmark:

https://github.com/tensorflow/benchmarks/tree/master/scripts/tf_cnn_benchmarks

python tf_cnn_benchmarks.py --num_gpus=1 --batch_size=32 --model=resnet50 --variable_update=parameter_server
One GPU GTX 750  27 GTX 960 48 GTX 1060 89