RStudio GPU Workstations in the Cloud with Paperspace

April 1, 2018
By

[This article was first published on TensorFlow for R, and kindly contributed to R-bloggers]. (You can report issue about the content on this page here)
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

We are very pleased to announce the availability of an RStudio TensorFlow template for the Paperspace cloud desktop service.

If you don’t have local access to a modern NVIDIA GPU, your best bet is typically to run GPU intensive training jobs in the cloud. Paperspace is a cloud service that provides access to a fully preconfigured Ubuntu 16.04 desktop environment equipped with a GPU. With the addition of the RStudio TensorFlow template you can now provision a ready to use RStudio TensorFlow w/ GPU workstation in just a few clicks. Preconfigured software includes:

  • RStudio Desktop and RStudio Server

  • NVIDIA GPU libraries (CUDA 8.0 and cuDNN 6.0)

  • TensorFlow v1.4 w/ GPU

  • The R keras, tfestimators, and tensorflow packages.

  • The tidyverse suite of packages (ggplot2, dplyr, tidyr, readr, etc.)

Getting Started

To get started, first signup for a Paperspace account (you can use the RSTUDIO promo code when you sign up to receive a $5 account credit).

Then, create a new Paperspace instance using the RStudio template:

Then, choose one of the Paperspace GPU instances (as opposed to the CPU instances). For example, here we select the P4000 machine type which includes an NVIDIA Quadro P4000 GPU:

See the Cloud Desktop GPUs with Paperspace article on the TensorFlow for R website for full details on getting started.

Training a Convolutional MNIST Model

The performance gains for training convoluational and recurrent models on GPUs can be substantial. Let’s try training the Keras MNIST CNN example on our new Paperspace instance:

Training the model for 12 epochs takes about 1 minute (~ 5 seconds per epoch). On the other hand, training the same model on CPU on a high end Macbook Pro takes 15 minutes! (~ 75 seconds per epoch). Using a Paperspace GPU yields a 15x performance gain in model training.

This model was trained on an NVIDIA Quadro P4000, which costs $0.40 per hour. Paperspace instances can be configured to automatically shut down after a period of inactivity to prevent accruing cloud charges when you aren’t actually using the machine.

If you are training convolutional or recurrent models and don’t currently have access to a local NVIDIA GPU, using RStudio on Paperspace is a great way to accelerate training performance. You can use the RSTUDIO promo code when you sign up for Paperspace to receive a $5 account credit.

To leave a comment for the author, please follow the link and comment on their blog: TensorFlow for R.

R-bloggers.com offers daily e-mail updates about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.



If you got this far, why not subscribe for updates from the site? Choose your flavor: e-mail, twitter, RSS, or facebook...

Comments are closed.

Search R-bloggers

Sponsors

Never miss an update!
Subscribe to R-bloggers to receive
e-mails with the latest R posts.
(You will not see this message again.)

Click here to close (This popup will not appear again)