Starting PottersWheel including installation
Copy the PottersWheel toolbox into your home folder at the cluster
using a Unix terminal, Cygwin or WinSCP
scp PottersWheel.zip username@server:
Start an X terminal (Cygwin: startx).
Log in:
ssh username@server -XC
Unzip the PottersWheel toolbox:
unzip PottersWheel.2.0.46.zip
Get a list of possible queues for your user account:
bqueues -u username
Start an interactive queue:
bsub -Is -q queueName bash
If no job is dispatched within a few minutes,
you can check for the reason using
bjobs -l
See a list of available Matlab licenses:
lmstat -f MATLAB_Distrib_Comp_Engine
Start Matlab:
matlab -nosplash -nodesktop
In Matlab, go into the toolbox folder and install PW:
cd PottersWheelToolbox
pwInstall
During installation, a warning might appear that the path definition
can not be changed. Then, the file pathdef.m will open in the Matlab
editor. Please save it into your home folder, if you start Matlab from there.
Go into a working directory, start PottersWheel,
load your model and data, and apply a fit
cd Modeling/Project01/WorkingFolder
pw
pwAddModel('M1.m');
pwAddData('Dataset1.txt');
pwCombine;
pwFit;
Alternatively, just use a model from the library
and simulate some data
pwAddModelFromLibrary('M1.m');
pwSim;
pwCombine;
Fit sequence
A fit sequence will be distributed to 20 nodes of the cluster.
Create a job using the distributed computing toolbox
jm = findResource('scheduler', 'type', 'lsf');
set(jm, 'ClusterMatlabRoot', '/opt/matlab');
job = createJob(jm);
set(jm, 'SubmitArguments', '-q queueName'); % Harvard:sysbio_15m,2h,12h,1d,7d,unlimite
Use the job within pwF2:
Syntax: pwF2(nFits, strength, backupMinutes, dataMode, tMax, job, nNodes)
Using e.g. nNodes = 20 and nFits = 100, 5 fits will be run on each node.
pwF2(100, [], [], [], [], job, 20);
