A small demo of how to use the Machine Learning Engine on Google Cloud Platform. This repo accompanies a presentation done for the Vancouver Machine Learning in Production Meetup group.
The presentation slides are available here.
Machine Learning Engine is a product on Google Cloud Platform that allows for highly managed training and deployment of Tensorflow models. Starting from a Tensorflow code that runs locally, I will show how to submit a distributed training job to the cloud, how to store the trained model to a GCP storage bucket, and how to use the hyperparameter optimization features. Finally, I will deploy the model to production. All this in front of a live audience! I will also go through the pros and cons of using ML-Engine and discuss how to use it with Keras+tensorflow models.
Dan Mazur is a senior data scientist at Grow Technologies where he is reinventing banking using data analytics and machine learning. Dan has a PhD in theoretical high energy astrophysics from UBC.