This site is old; please update your bookmark to https://cbiit.github.com/sdsi/candle!
This page will (1) show you how to run a sample job using CANDLE and (2) indicate what needs to be modified in order to run your own job using CANDLE.
Step (1): Set up your environment
First, enter some working directory on Biowulf and load the CANDLE environment using module load candle
. For example, if your working directory is /data/USERNAME/candle-test
, run this on the Biowulf command line:
cd /data/USERNAME/candle-test
module load candle
Step (2): Copy the template submission script to the working directory
Copy the template CANDLE submission script submit_candle_job.sh
to, and create an empty experiments
directory in, the working directory by running on Biowulf:
copy_candle_template
Step (3) (Optional): Modify the variables in the submission script
Modify the template submission script submit_candle_job.sh
to your needs or leave it unmodified for an example run on the MNIST dataset.
Step (4): Run the job
Submit the job by running:
./submit_candle_job.sh
(No, there really is no need for “sbatch”.)
Notes
- Feel free to email Andrew Weisman with any questions.