R Interface to Google CloudML

Overview

We are excited to announce the availability of the cloudml package, which provides an R interface to Google Cloud Machine Learning Engine. CloudML provides a number of services including:

• Scalable training of models built with the keras, tfestimators, and tensorflow R packages.

• On-demand access to training on GPUs, including the new Tesla P100 GPUs from NVIDIA®.

• Hyperparameter tuning to optmize key attributes of model architectures in order to maximize predictive accuracy.

• Deployment of trained models to the Google global prediction platform that can support thousands of users and TBs of data.

Training with CloudML

Once you’ve configured your system to publish to CloudML, training a model is as straightforward as calling the cloudml_train() function:

library(cloudml)
cloudml_train("train.R")

CloudML provides a variety of GPU configurations, which can be easily selected when calling cloudml_train(). For example, the following would train the same model as above but with a Tesla K80 GPU:

cloudml_train("train.R", master_type = "standard_gpu")

To train using a Tesla P100 GPU you would specify "standard_p100":

cloudml_train("train.R", master_type = "standard_p100")

When training completes the job is collected and a training run report is displayed:

Learning More

Check out the cloudml package documentation to get started with training and deploying models on CloudML.

You can also find out more about the various capabilities of CloudML in these articles:

• Training with CloudML goes into additional depth on managing training jobs and their output.

• Hyperparameter Tuning explores how you can improve the performance of your models by running many trials with distinct hyperparameters (e.g. number and size of layers) to determine their optimal values.

• Google Cloud Storage provides information on copying data between your local machine and Google Storage and also describes how to use data within Google Storage during training.

• Deploying Models describes how to deploy trained models and generate predictions from them.