setup.txt

Install Anaconda (Python 3.6 version)
https://www.anaconda.com/download/#linux

Install CUDA:
http://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html

Install keras and tensorflow-gpu backend (GPU backend for efficient training):
conda update --all
conda install tensorflow-gpu keras

If you ONLY want to do inference (i.e. make predictions) the CPU backend is sufficient:
conda install tensorflow keras


The Tensorflow notebook works without any further modifications.

Edit ~./keras/keras.json only if using the Theano backend:

{
    "epsilon": 1e-07,
    "backend": "theano",
    "floatx": "float32",
    "image_dim_ordering": "th",
    "image_data_format": "channels_first"
}


run nvidia-smi to check CUDA status:
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 387.26                 Driver Version: 387.26                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 1070    Off  | 00000000:09:00.0  On |                  N/A |
|  0%   45C    P2    64W / 200W |   7883MiB /  8105MiB |     98%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      1264      G   /usr/lib/xorg/Xorg                           110MiB |
|    0      4747      C   /home/js/anaconda3/bin/python               7685MiB |
+-----------------------------------------------------------------------------+

Open the Jupyter notebook:
jupyter notebook emnist.ipynb               # when using the theano backend
jupyter notebook emnist_tensorflow.ipynb    # when using the tensorflow backend