Issues installing INLA + fmesher on HPC (Ginsburg)
Ji Gna
Hi,
I’m trying to set up INLA with fmesher on the Columbia University Ginsburg HPC (R/4.5.1, GCC 13.0.1).
What I’ve done / issues encountered:
-
Installed INLA with INLA:::inla.binary.install() → installation succeeded.
-
library(INLA) fails with:
Error: package or namespace load failed for 'INLA': there is no package called 'fmesher' -
Installing fmesher fails because sf requires s2.
-
Building s2 from source did not succeed (system-level dependencies seem missing).
-
The cluster R/4.5.1 module was built without --enable-R-shlib, so libR.so is not available; this caused earlier package load errors.
-
Tried using a conda R 4.5.1 environment (with udunits2, GDAL, PROJ, GEOS), but still cannot get s2/sf/fmesher installed.
-
Tried disabling renv (RENV_CONFIG_AUTOLOAD=FALSE) to avoid old libraries, but the dependency chain problem remains.
At this point, INLA is installed but cannot load without fmesher, and fmesher cannot be installed because sf/s2 won’t build.
Does anyone have suggestions on how to solve this?
Thanks,
Finn Lindgren
the proper solution I think is to work with your system administrators to install the missing system requirements for s2. Every HPC system is different, and they should be best placed to know what needs to be done, and how.
What type of models are you intending to run. Anything involving spde/mesh models would require fmesher and sf.
It might be possible to modify the INLA code to not need a hard dependency on fmesher, so that non-spde models could be run without installing fmesher, but that would require work.
The fmesher code itself might not strictly require sf, but it might make use of some sf methods even when the input data isn't in sf format, so I'm not sure it would be fully usable without sf at the moment. And even if it is possible to drop the hard dependency on sf to a Suggests, it would take work to do so.
As for s2 itself, INLA/fmesher doesn't use it if we can avoid it, but I don't think there's a way to disentangle it from sf.
Finn
--
You received this message because you are subscribed to the Google Groups "R-inla discussion group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to r-inla-discussion...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/r-inla-discussion-group/ac2601a1-ce5d-4028-b022-dcc9b0547377n%40googlegroups.com.