Features

This is a data science playground. It includes:

• Jupyter Notebook.
  • Supports Python 3.6 and Python 2.7 as different Jupyter Notebook kernels.
  • Accessible at localhost:8888.
  • The root directory of the notebook is analysis.
  • Supports all the libraries tensorflow-notebook Docker image supports, for both versions of Python.
• Tensorboard support.
  • Accessible at localhost:6006.
  • Reads logs from analysis/logs.
  • Exposes TENSORBOARD_LOG_DIR environmental variable that points to that directory within the Docker container. You can use this variable from the notebooks.
• Postgres database in a separate Docker container, accessible at localhost:5432.
  • You can initialize it with custom SQL scripts by placing them in the database folder and uncommenting the ADD ./*.sql /docker-entrypoint-initdb.d/ line in database/Dockerfile.
• Demo notebooks with examples of the notebook usage.
  • Keras & Tensorflow: demonstrates how to use these two with Tensorboard that ships with this playground.
  • Postgres: demonstrates how to connect to the Postgres database of this sandbox and read data as Pandas DataFrames.

Running the Servers

To run the playground, execute the following command:

docker-compose down; docker-compose build; docker-compose up

You will get the above features accessible at the specified URLs.

To access Jupyter Notebook, you will need an access token. Keep an eye on the console output of the above command and look for the lines like follows:

tfpg_analysis | [I 14:33:07.656 NotebookApp] The Jupyter Notebook is running at:
tfpg_analysis | [I 14:33:07.656 NotebookApp] http://(42140b2eee79 or 127.0.0.1):8888/?token=e2ab752a6e907b44f28f72dfd9e7c6e39957051e5b802753

The token here is: e2ab752a6e907b44f28f72dfd9e7c6e39957051e5b802753.

Architecture

The playground consists of two docker containers:

• tfpg_analysis – contains the Jupyter Notebook, TensorBoard and many Python libraries for Data Science.
• tfpg_database – contains the Postgres database.

tfpg_database persists its data under the _volumes/postgres-data directory. The analysis directory is mounted to the home directory of the tfpg_analysis container default user – so it can be accessed from that container.

Connecting to the Analysis server

You can open a command line shell at the tfpg_analysis container to run Python scripts placed under analysis directory. To do so, after you start the servers as described in "Running the Servers", run:

docker exec -ti tfpg_analysis start.sh

If you would like to be able to use passwordless sudo from the container, run the following command:

docker exec -tiu root tfpg_analysis start.sh