GPU Workstations within the Cloud with Paperspace

GPU Workstations within the Cloud with Paperspace

We’re very happy to announce the provision of an RStudio TensorFlow template for the Paperspace cloud desktop service.

Should you don’t have native entry to a contemporary NVIDIA GPU, your greatest wager is often to run GPU intensive coaching jobs within the cloud. Paperspace is a cloud service that gives entry to a completely preconfigured Ubuntu 16.04 desktop atmosphere geared up with a GPU. With the addition of the RStudio TensorFlow template now you can provision a prepared to make use of RStudio TensorFlow w/ GPU workstation in only a few clicks. Preconfigured software program consists of:

  • 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, and so on.)

Getting Began

To get began, first signup for a Paperspace account (you should utilize the RSTUDIO promo code if you signal as much as obtain a $5 account credit score).

Then, create a brand new Paperspace occasion utilizing the RStudio template:

Then, select one of many Paperspace GPU situations (versus the CPU situations). For instance, right here we choose the P4000 machine sort which incorporates an NVIDIA Quadro P4000 GPU:

See the Cloud Desktop GPUs with Paperspace article on the TensorFlow for R web site for full particulars on getting began.

Coaching a Convolutional MNIST Mannequin

The efficiency features for coaching convoluational and recurrent fashions on GPUs could be substantial. Let’s attempt coaching the Keras MNIST CNN instance on our new Paperspace occasion:

Coaching the mannequin for 12 epochs takes about 1 minute (~ 5 seconds per epoch). Then again, coaching the identical mannequin on CPU on a excessive finish Macbook Professional takes quarter-hour! (~ 75 seconds per epoch). Utilizing a Paperspace GPU yields a 15x efficiency acquire in mannequin coaching.

This mannequin was educated on an NVIDIA Quadro P4000, which prices $0.40 per hour. Paperspace situations could be configured to mechanically shut down after a interval of inactivity to stop accruing cloud expenses if you aren’t really utilizing the machine.

If you’re coaching convolutional or recurrent fashions and don’t at present have entry to a neighborhood NVIDIA GPU, utilizing RStudio on Paperspace is an effective way to speed up coaching efficiency. You need to use the RSTUDIO promo code if you join Paperspace to obtain a $5 account credit score.


For attribution, please cite this work as

Allaire (2018, April 2). RStudio AI Weblog: GPU Workstations within the Cloud with Paperspace. Retrieved from

BibTeX quotation

  creator = {Allaire, J.J.},
  title = {RStudio AI Weblog: GPU Workstations within the Cloud with Paperspace},
  url = {},
  12 months = {2018}


Please enter your comment!
Please enter your name here