Dockerized Jupyter with Keras.
With port forwarding:
docker run -d -p 8888:8888 ermaker/keras-jupyter
With Tensorflow: (See this for more information)
docker run -d -p 8888:8888 -e KERAS_BACKEND=tensorflow ermaker/keras-jupyter
For persistent storage:
docker run -d -p 8888:8888 -v /notebook:/notebook ermaker/keras-jupyter
Just browse localhost:8888 and write code with Keras!