How do I use the RStudio Cloud for teaching Keras and TidyVerse?

tidyverse
keras
rstudiocloud
tensorflow

#1

Dear fellow R-aficionados,

I am to teach a deep learning class to ~30 students at university level. I have a set of IPython Notebook lying around, but I'd naturally prefer using R, because... R FTW :smiley:

I have really bad experiences with having students install tidyverse and keras on their laptops, so cloud here I come, but how to?

Optimally, I would like to have a completely prepared workspace, with tidyverse, keras and then a set of prepared scripts and data, which the students can then access after creating an account.

Is this feasible? Not only access-wise, but also computational power given, that we will be using tensorflow?

Thanks in advance,
Leon


#2

I am considering the same thing. @mine has some excellent resources on Rstudio Cloud. See response here: Equivalent of JupyterHub with Rstudio (Cloud?)


#3

I'm not sure about RStudio Cloud, but I've recommended this software elsewhere. Combined with Docker it can probably make student's lives a bit easier? Although, of course, installing Docker can be a problem in and of itself...


#4

The only issue you might have with RStudio Cloud is the resource requirements (as I think the instances you get there currently don't have GPUs and have ~ 2 GB of RAM).

Another thought would be to just install RStudio Server or RStudio Server Pro (which is free for teaching, see https://www.rstudio.com/pricing/academic-pricing/)


#5

Hi @jjallaire,

Nice to talk to you again :slight_smile:

I have definitely been thinking about RStudio-server, but due to time constraints, my plan is to prepare all needed files on my own RStudio Cloud account and then let each student create their own account and then let them upload the file portfolio I am creating as a zip-file.

I myself seamlessly installed tidyverse and keras just now, which I will have them do and then combined with the files I am working on, we should be good to go...

I have ~30 students - I hope that's not too much given your resources?

...and thanks for making RStudio Cloud available and free-of-use - That's really awesome!
Bw. Leon


#6

Just to reinforce what JJ said, rstudio.cloud is currently limited to 1GB of memory and does not have access to a GPU.

If that is sufficient for what you need to accomplish in the class, wonderful. You could look into creating a public project that the students could copy that contains the files and installed R packages, rather than having them upload the same zip-file, or using a Space to organize the students and their projects. The number of students is not an issue.

While we expect to make the option of more memory and a GPU available in the future, I would strongly suggest running through the coursework to ensure that it stays within the memory limit, otherwise the students will have a subpar experience.

Hope that helps.


#7

Thanks for input - I am running the setup now and everything seems to be going neatly.

I am not going to have them build any massive models, it's very basic, so I hope (and it is the case so far) I will all good with 1-gig...

As I am a bit under time pressure, for now I just need something that works and then later on I will explore more elegant solutions :slight_smile:


#8

Everything went really well with the class today - Installing and running tidyverse and keras on the rstudio.cloud solution worked seamlessly and there was enough ram and diskspace for the exercises! :+1:


#9

I highly recommend creating a space and template project for your students, so they don't need to install tidyverse and keras as their first activity, unless that's part of what you want to teach them.