This sample uses the Google Cloud Machine Learning API, Tensorflow, Apache Beam, and the provided Cloud ML SDK to,

• Preprocess data
• Start a Cloud ML API Training Job
• Create a Cloud ML model, from the output
• Start a Cloud ML Prediction service
• Predict values of new instances.

To run this example, first follow instructions for setting up your environment, and preprocess the data using preprocess.py, then you may follow instructions in the MNIST quickstarts replacing relevant filepaths with ones which refer to this directory.

Alternatively you may run the example end-to-end in Dataflow. Run,

python pipeline.py --help

for the full list of required and optional command line arguments.