Tensorflow Tutorial

Example with Face Embeddings

Setting up an environment

Clone this repo and cd into it.

git clone https://github.com/abaybektursun/tf_tutorial
cd tf_tutorial

Make sure that you have virutalenv

pip3 install virtualenv

Now we will create a virtualenv. Follow the steps for your operating system

Linux

virtualenv --system-site-packages -p python3 ./venv
source ./venv/bin/activate
pip3 install --upgrade pip

Windows

virtualenv --system-site-packages -p python3 ./venv
.\venv\Scripts\activate

Installing Tensorflow and the other dependencies

If you have an NVIDIA card and want it to be utilized by TF, you can follow the instructions for GPU. Otherwise follow the CPU instructions.

CPU

pip3 install -r requirements_cpu.txt

GPU

pip3 install -r requirements_gpu.txt

In addition, you will have to have CUDA 9.0 and cuDNN installed.

• Install CUDA 9.0 from here: https://developer.nvidia.com/cuda-90-download-archive
• Install cuDNN by following instructions for your specific platform here. Make sure to downlad the cuDNN specifically for CUDA 9.0: https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html
• On Windows, GPU version of Tensorflow only supports Python 3.5.x

Jupyter Notebook

Install new kernel for the jupyter notebook

ipython kernel install --user --name=tf_tutorial

Star the jupyter notebook

jupyter notebook