Workshop / Examples Container

This is a Docker image for running the TFX Developer Tutorial. It includes TensorFlow and TFX, and initializes a clean, basic environment for running the workshop.

Prequisites

• Mac or Linux (Highly recommended)
• Windows (optional)
• Docker
• Git
• At least 3GB available disk space

To Run

git clone https://github.com/tensorflow/workshops.git
cd tfx_airflow
source start_container.sh

Note: Instructions for Windows are in windows_notes.md.

Running the TFX Developer Tutorial

Follow the instructions in the tutorial, which assume that you are running on your host, and make a few changes to commands which are required because you are running inside a container instead of directly on your machine.

Do not start Airflow webserver, Airflow scheduler, or Jupyter notebook. Those have already been started for you in your container.

The Airflow console will be available on the host system (in other words, outside the container) at:

http://localhost:8080

Jupyter Notebook will be available on the host system at:

http://localhost:8888

Using Bash (optional)

Once the container startup is complete you will be in a bash shell, which will be logging Airflow messages. Hit return to get to a bash prompt, or start another shell in a different terminal. To run another shell, move to where you cloned the workshops repo and run:

source run_bash.sh

Since the workshop runs in a Python virtual environment you may also need to activate that environment in the bash shell that you're running inside the container. It is located in the container at /root/tfx_env.

source /root/tfx_env/bin/activate

To exit your shells, enter exit.