Beagle install
Installing beagle from source can be a bit of a pain. Installing & compiling from source is difficult and always causes problems.
TODO: Explain how to install from source on HPC
Cheat with Singularity
what is Singularity?
Ooh you don't know what singularity is? Have a read here
The easiest way to use beagle is through a container. The container is a self-contained environment that has all the dependencies installed.
A central container repository has multiple versions of the BEAST-BEAGLE software that is compatible with CUDA (GPU).
https://quay.io/repository/broadinstitute/beast-beagle-cuda
But these are limited to their repo. I've made a couple of modified containers for a phylogenetic factor analysis and a latent variable model. Which can be found on the dockerhub: - Beast & latent-variable - PhylogeneticFactorAnalysis.jl
To use one of these containers, simply run:
module load CUDA
singularity exec --nv \
-B $VSC_DATA:$VSC_DATA -B $VSC_STAGING:$VSC_STAGING -B $VSC_SCRATCH:$VSC_SCRATCH -B $VSC_HOME:$VSC_HOME -B $PWD \
docker://quay.io/broadinstitute/beast-beagle-cuda:v10.5.0_beta3 \
beast -beagle_info
Tip
All these -B statements are to bind the directories to the container. Meaning, the container can access the directories on the host machine.
Tip
The --nv flag is to enable GPU support.
Tip
we specified docker:// because the container is stored in a docker repository.