Running

Running

Blue Waters

Get a python REPL on Blue Waters compute node

aprun -b -n1 -- bwpy-environ -- python -c "import code;code.interact()"

Single-Node Test

Get an interactive node

qsub -I -l nodes=1:ppn=16:xk -l walltime=00:30:00

Load bwpy / bwpy-mpi

module load bwpy
module load bwpy-mpi

Preprocessing output:

aprun -b -n1 -- bwpy-environ -- python-mpi examples/guarantee_preprocessed.py

5522/5523Finished Preprocessing 5523 files in 51303.606929302216 seconds
Omitted 22 shots of 5523 total.
479/5501 disruptive shots
4327 1196
validate: 1424 shots, 90 disruptive
training: 2893 shots, 215 disruptive
testing: 1184 shots, 174 disruptive
...done
Application 67642076 resources: utime ~50941s, stime ~382s, Rss ~290960, inblocks ~124329092, outblocks ~27902089

One of the tests doesn't work in Python 3.5. Unsure whether this really a problem yet.

EPYTHON="python2.7" aprun -b -n1 -- bwpy-environ -- python-mpi test.py

Getting the data:

The data is available at the Princeton TIGRESS FRNN Globus Endpoint

Modify examples/conf.yaml to aim at whereever you put the data. For example, if the data is in ~/plasma-data, then:

fs_path: '/mnt/a/u/sciteam'
paths.signal_prepath: '/plasma-data/signal_data/'
paths.shot_list_dir: '/plasma-data/shot_lists/'

Learning

Get 4 interactive nodes

qsub -I -l nodes=4:ppn=16:xk -l walltime=00:30:00

Even though one test didn't work in python3.5, try python3.5.

module load bwpy bwpy-mpi

aprun -b -n1 -- bwpy-environ -- python-mpi learn.py