GPU workstations in the cloud with Paperspace

We are very happy to announce the availability of one. RStudio TensorFlow template for the Paper space Cloud Desktop Service.

If you don't have local access to a modern NVIDIA GPU, your best bet is usually 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 RStudio TensorFlow template Now you can provision a ready-to-use RStudio TensorFlow w/ GPU workstation in just a few clicks. Pre-configured software includes:

• RStudio Desktop and RStudio Server

• NVIDIA GPU Libraries (CUDA 8.0 and cuDNN 6.0)

• TensorFlow v1.4 w/ GPU

• R Kira, tfestimatorsAnd Tensor flow Packages

• gave neat and clean A collection of packages (ggplot2, dplyr, tidyr, readr, etc.)

getting started

To begin with, first Sign up for a Paperspace account. (You can use RSTUDIO promo code when you sign up to get a $5 account credit).

Then, create a new PaperSpace instance using R Studio Template:

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

See Cloud Desktop GPUs with Paperspace Article on the TensorFlow for R website for complete details on how to get started.

Training a Convolutional MNIST Model

The performance gains can be substantial for training continuous and iterative models on GPUs. Let's try to train. Keras MNIST CNN example On our new Paperspace example:

It takes about 1 minute (~5 seconds per epoch) to train the model for 12 epochs. On the other hand, training the same model on a high-end Macbook Pro on CPU takes 15 minutes! (~75 seconds per round). Using Paperspace GPU a 15x performance increase In model training

The model was trained on a NVIDIA Quadro P4000That costs $0.40 an hour. PaperSpace instances can be configured to automatically shut down after a period of inactivity to prevent cloud charges from accruing when you're not actually using the machine.

If you are training a convolutional or recurrent model and do not currently have access to a native NVIDIA GPU, R Studio on Paperspace It is a great way to speed up training performance. You can use RSTUDIO Promo code when you sign up for Paperspace to get $5 account credit.

Reference

For attribution, please cite this work.

Allaire (2018, April 2). Posit AI Blog: GPU Workstations in the Cloud with Paperspace. Retrieved from 

BibTeX reference

@misc{allaire2018gpu,
  author = {Allaire, J.J.},
  title = {Posit AI Blog: GPU Workstations in the Cloud with Paperspace},
  url = {},
  year = {2018}
}