Skip to content

Latest commit

 

History

History
95 lines (70 loc) · 3.47 KB

INSTALL.md

File metadata and controls

95 lines (70 loc) · 3.47 KB

OSG-distributed conda environments

If you have access to OSG containers via CVMFS, you can use pre-assembled IGWN conda install

source /cvmfs/oasis.opensciencegrid.org/ligo/sw/conda/bin/activate 
conda activate  igwn-py37-testing  # omit 'testing' in production
export GW_SURROGATE=''
export LIGO_USER_NAME=albert.einstein
export LIGO_ACCOUNTING=ligo.dev.o3.cbc.pe.lalinferencerapid

Full installation

You can install RIFT via pip (and conda). If you are a developer you will need to install from source.

  • pip install: Installation with pip is the easiest.
  pip install  --user  RIFT
  • source install with setup.py: If you retrieve the source code, you can run the setup script directly. That's very helpful if you need to edit the source
git clone https://github.com/oshaughn/research-projects-RIT.git # use for HTTPS
cd research-projects-RIT
python setup.py install --user
export GW_SURROGATE=''
  • Alternatively you can use pip install --user -e . from the source directory
  • source install, directly : Finally, if you don't want to use pip, you can work directly with the source
git clone https://github.com/oshaughn/research-projects-RIT.git # use for HTTPS
# git clone [email protected]:oshaughn/research-projects-RIT.git # use instead for SSH
cd research-projects-RIT
git checkout temp-RIT-Tides-port_master-GPUIntegration   
export INSTALL_DIR=`pwd`
export GW_SURROGATE=''
  * put the following directory in your `PATH` and `PYTHONPATH`

export ILE_DIR=${INSTALL_DIR}/MonteCarloMarginalizeCode/Code
export PATH=${PATH}:${ILE_DIR}
export PYTHONPATH=${PYTHONPATH}:${ILE_DIR}
export GW_SURROGATE=''
*  make sure you have installed the necessary dependencies.  The following command will at least ensure that these dependencies are up to date 

python setup.py install --user

lalsuite

The code relies heavily on lalsuite. Please have it installed and working properly.

The code also requires a working version of glue, supporting glue.ligolw.ligolw. The pip installable version of glue has diverged from the version installed on LDG clusters, and this part of glue has moved into lalsuite. We'll update accordingly, but be warned that ongoing rapid glue evolution may cause you version-matching headaches (e.g., if the XML format changes).

GPU-sensitive installs

The code uses cupy to access GPUs. If you don't have one, the code will still work. If you do need one, make sure to install cupy **on a machine that supports GPUs **

pip install --user cupy

Other environment variables

If you run on an LDG cluster, you need accounting flags

export LIGO_USER_NAME=albert.einstein
export LIGO_ACCOUNTING=ligo.dev.o3.cbc.pe.lalinferencerapid

Alternative setup (singularity)

If you would rather use a pre-packaged environment and you have access to singularity and CVMFS (e.g., on an LDG cluster), you can do the following to use a pre-packaged version :

singularity shell --writable  /cvmfs/ligo-containers.opensciencegrid.org/james-clark/research-projects-rit/rift/latest

You can use this setup for testing and job launching.

Alternative setup (docker)

You can also use docker to retrieve a docker image maintained by James Clark docker hub location

docker pull jclarkastro/rift
docker run --detach --gpus all -v /home:/home -v /cvmfs:/cvmfs --name=rift-demo jclarkastro/rift