GPU@RStudio for teaching TensorFlow/Keras at University - Solution?

For the next iteration of my course with ~50 students, I would like to dive more into TensorFlow/Keras for R.

I am therefore looking for suggested solutions for GPU@RStudio for teaching TensorFlow/Keras at University? Amazon AWS? Google Cloud? Other?

Currently we have simply been running on, which is quite limited in terms of hardware ressources.

Thank you,

1 Like

Current free options always involves notebooks:

The problem with setting your own cloud based solution is that TensorFlow takes a lot of GPU memory, so basically you would need 1 GPU for each student and this gets very fast unaffordable.

Paperspace seems to be a possible solution:
You should be able to setup RStudio in it:

Excellent input - Thanks!

So... Where do I rent ~50 GPU instances with RStudio for teaching for 8-10 hours per week for a total of 13 weeks and what should I expect that it would cost?

Nah bro, get your students to make kaggle accounts and they'll get free GPU time. I think it's like 3 hours a day?

If you decide to rent server instance from aws google cloud, you will have tons of configuration setup management work for each of student environment. Use our server all you need is 3 button click.... :slight_smile: check out details here: the cost will be much less than other vendor. Leave us a message I will create demo account for you.

My recommendation is to create a GPU kubernetes cluster with GPU and deploy Rstudio containers with GPU support. This will be a lot cheaper. GCP is the easiest solution. Rocker/rstudio already has GPU supported rstudio container.

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.