How to install Python3.6 and use Tensorflow VGG16 example.

Welcome to the cifar10-vgg16 wiki!

Its a simple tutorial to show how to run this code with GPU. There are lots of missing parts, you can open issues. Im using Windows7 64bit ultimate. To run the code these packages needed to be installed.

• install python 3.6.7 from python page.
• install CUDA 9.0
• download cudnn 7.0.5 base installer and latest patch5( now the lates is patch5 at 11.18.2018). You need to register nvidia developer account.

after installing cuda9.0, there is "cudnn64_7.dll" in the downloaded cudnn7.0.5 package inside bin folder. Copy this dll to "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0\bin" directory.

Open windows cmd as administrator!!!

• pip install tensorflow-gpu==1.5.0
• pip install Keras==2.2.0
• pip install spyder # spyder ide fore for debugging python code. Very useful!

Open spyder3 for searching spyder3 in windows search(press windows button in keyboard and search it) if any of example is not working try this. (i received this errors in spyder3 ide: AttributeError: module 'tensorflow' has no attribute 'name_scope') Reinstall as;

• pip install tensorflow-gpu==1.5.0
• pip install Keras==2.2.0

#tried this solution from stackoverflow solutions.

1. pip install --upgrade pip setuptools wheel
2. pip install -I tensorflow
3. pip install -I keras

if you need to work with opencv please install specific version of opencv instead of opencv-3.4.3.1. See Why

• pip install opencv-python==3.4.1.15

There should be lots of packages which are not installed. You can install them such like pip install "missingpackage"; For example: pip install matplotlib

Thanks.