R Interface to Google CloudML

We’re excited to announce the provision of the cloudml bundle, which supplies an R interface to Google Cloud Machine Studying Engine. CloudML supplies plenty of providers together with on-demand entry to coaching on GPUs and hyperparameter tuning to optimize key attributes of mannequin architectures.

Overview

We’re excited to announce the provision of the cloudml bundle, which supplies an R interface to Google Cloud Machine Learning Engine. CloudML supplies plenty of providers together with:

• Scalable coaching of fashions constructed with the keras, tfestimators, and tensorflow R packages.

• On-demand entry to coaching on GPUs, together with the brand new Tesla P100 GPUs from NVIDIA®.

• Hyperparameter tuning to optmize key attributes of mannequin architectures with a view to maximize predictive accuracy.

• Deployment of skilled fashions to the Google international prediction platform that may help hundreds of customers and TBs of information.

Coaching with CloudML

When you’ve configured your system to publish to CloudML, coaching a mannequin is as easy as calling the cloudml_train() operate:

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

CloudML supplies a wide range of GPU configurations, which could be simply chosen when calling cloudml_train(). For instance, the next would prepare the identical mannequin as above however with a Tesla K80 GPU:

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

To coach utilizing a Tesla P100 GPU you’ll specify "standard_p100":

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

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

Studying Extra

Try the cloudml package documentation to get began with coaching and deploying fashions on CloudML.

You can too discover out extra concerning the varied capabilities of CloudML in these articles:

• Training with CloudML goes into extra depth on managing coaching jobs and their output.

• Hyperparameter Tuning explores how one can enhance the efficiency of your fashions by operating many trials with distinct hyperparameters (e.g. quantity and dimension of layers) to find out their optimum values.

• Google Cloud Storage supplies data on copying knowledge between your native machine and Google Storage and in addition describes how one can use knowledge inside Google Storage throughout coaching.

• Deploying Models describes how one can deploy skilled fashions and generate predictions from them.

Reuse

Textual content and figures are licensed beneath Inventive Commons Attribution CC BY 4.0. The figures which were reused from different sources do not fall beneath this license and could be acknowledged by a be aware of their caption: “Determine from …”.

Quotation

For attribution, please cite this work as

Allaire (2018, Jan. 10). Posit AI Weblog: R Interface to Google CloudML. Retrieved from https://blogs.rstudio.com/tensorflow/posts/2018-01-10-r-interface-to-cloudml/

BibTeX quotation

@misc{allaire2018r,
  creator = {Allaire, J.J.},
  title = {Posit AI Weblog: R Interface to Google CloudML},
  url = {https://blogs.rstudio.com/tensorflow/posts/2018-01-10-r-interface-to-cloudml/},
  12 months = {2018}
}