This Dockerfile builds an image containing both Tensorflow and Theano, but enables TF as a backend for Keras.
You can add any files you like to this project and rebuild the image to experiment with them.
On OSX, you will need to install Docker Machine or boot2docker before running this image.
docker build . --tag peterrosell/keras-jupyter
-
Make a directory named
notebooks
in your home directory. -
Run the following command from command line
docker run \
-it \
--rm \
--name keras \
-e KERAS_BACKEND=tensorflow \
-p 8888:8888 \
-p 6006:6006 \
-v ${HOME}/notebooks:/notebooks \
peterrosell/keras-jupyter
-
Browse to http://localhost:8888 to access the GUI for jupyter Notebooks. Browse to http://localhost:6006 to access Tensorboard. Tensorboard is used for displaying log results from a learing session.
-
To shutdown just press Ctrl-C at command line.