cloudml-magic is a Jupyter Notebook Magics for interactively working with Google Cloud Machine Learning Engine. Your Tensorflow or Keras code on Notebook will run on ML Engine, just by running cells.
Before you begin, prepare Google Cloud Platform project, enable billing and install Google Cloud SDK.
Also this magics use Application Default Credentials. To activate the credentials, enter
gcloud auth application-default login
To install this magics
pip install cloudmlmagic
To use this magics, enter the following command in your notebook.
in [1]: %load_ext cloudmlmagic
That's it!
This magic initializes your cloud ml job request. Do NOT forget to initialize before adding code or run.
Example:
in [2]: %ml_init -projectId PROJECTID -bucket BUCKET -region us-central1 -scaleTier BASIC -runtimeVersion 1.2
If you want to run codes with external libraries like Keras, h5py, add following dict below the %ml_init
magic.
{'install_requires': ['keras', 'h5py', 'Pillow']}
This magic adds the block into uploading package.
To run a training job, add this magic on your code block.
Run on local Machine
Do not add this magic, or add %ml_run
Run on Cloud ML Engine
add %ml_run cloud