How-to run R

Versions

As of this article update, we have the following versions of R installed on Talapas:

3.3.3
3.3.3_intel-17
3.4.1
3.4.2
3.4.3
3.4.4
3.5.1
3.5.3
3.6.1
3.6.1_gcc7
3.6.2
4.0.2
4.0.2_gcc7

Note: packages within each version of R are not the same! There may be different packages or different package versions.

Intel R

Intel is the default. To run an Intel compiled version of R, e.g 4.0.2, run:

$ module load R/4.0.2
$ R

gcc R

We have 2 versions of R compiled with gcc7 (3.6.1_gcc7 and 4.0.2_gcc7) built specifically to support RStan. To use these versions you must first load gcc. E.g. 4.0.2, run:

$ module load gcc

Lmod is automatically replacing "intel/17" with "gcc/7.3".

$ module load R/4.0.2
$ R

Verify the path of the R executable that is loaded use the bash command which:

$ which R
/packages/R/4.0.2_gcc7/bin/R

The output above shows the executable to R version 4.0.2 built with gcc7.

R Packages

To list installed packages and version information, run:

$ module load R/4.0.2
$ R
> installed.packages()[,c("Package","Version")]
                     Package                Version
abind                "abind"                "1.4-5"
...

Running R Interactive

ssh onto a talapas login node, https://hpcrcf.atlassian.net/wiki/spaces/TCP/pages/6762250
Start an interactive job, https://hpcrcf.atlassian.net/wiki/spaces/TCP/pages/7089568
Once on the allocated node enter,

$ module load R/4.0.2
$ R

Running R in Batch

For longer running non-interactive use cases call R in batch.

First, create your file, e.g. demo.r, which contains R syntax.

Next, create your batch job file, e.g. r-job.srun, and add desired slurm declarations, module(s), and call to the application.

#!/bin/bash

#SBATCH --account=<pirg-name>
#SBATCH --job-name=r-job
#SBATCH --output=r-job.out
#SBATCH --error=r-job.err

module load R/4.0.2

R --save < /gpfs/projects/<pirg-name>/<user-name>/demo.r

‘Save current workspace’ with --save

The example above preserves R session output. Also include the save.image function in your demo.r file, e.g. save.image("/gpfs/projects/<pirg-name>/<user-name>/demo.RData")

Submit the job

Use the sbatch command to submit the job.

$ sbatch r-job.srun