This is the main home page about CANDLE Supervisor effort with links to workflows and other supporting information.

Workflows

The workflows are currently indexed in the README visible here.

Database integration

The database work is described in the README visible here.

Swift installations

Theta

This is linked to Python and R but currently without ML libs.

Other Theta ESP notes are here: https://collab.cels.anl.gov/display/ESP

Python

Installed in:

/projects/Candle_ECP/swift/deps/Python-2.7.12

To run this installation, you must set:

$ export LD_LIBRARY_PATH=/projects/Candle_ECP/swift/deps/Python-2.7.12/lib

  • Cori

    This uses the system-installed Python with ML libs in module:
    tensorflow/intel-head

  • Titan

    This is a CANDLE-only installation. It uses the OLCF-provided Python deeplearning module (Python 3.6 plus TensorFlow, Theano, and Keras) and R 3.3.2 .

    Add to PATH: /lustre/atlas2/csc249/proj-shared/sfw/swift-t/stc/bin

    Run with:

    $ export TITAN=true
$ export PROJECT=... QUEUE=...
$ export LD_LIBRARY_PATH=/sw/xk6/deeplearning/1.0/sles11.3_gnu4.9.3/lib:/sw/xk6/deeplearning/1.0/sles11.3_gnu4.9.3/cuda/lib64:/opt/gcc/4.9.3/snos/lib64:/sw/xk6/r/3.3.2/sles11.3_gnu4.9.3x/lib64/R/lib
$ swift-t -m cray -e LD_LIBRARY_PATH=$LD_LIBRARY_PATH workflow.swift

On Titan, do module load autoconf to get Autoconf 2.69 .

  • Cooley

    This uses the system-installed Python with ML libs at:
    /soft/analytics/conda/env/Candle_ML

  • JLSE KNL

    This does not yet have Python.

  • JLSE Prasanna
    This uses a VirtualEnv Python at /home/pbalapra/.virtualenvs

    • ~wozniak/Public/sfw/icc/swift-t-pb/stc/bin