Praxis Verbose

Using the Multiple Run Fitter, praxis, etc..
Post Reply
roybens
Posts: 44
Joined: Fri Mar 14, 2008 7:57 am

Praxis Verbose

Post by roybens » Wed Dec 02, 2009 5:28 am

Hello
I am trying to run multiple runs of praxis using mpi, the problem is that I can't cancel printing using printmode= 0 at attr_praxis function and the logfiles gets HUGE and causing problems in our cluster.
Am I using the function correctly? Is there any other way to omit the output to the screen? I don't mind omitting all NEURON output.
Thanks
Roy

ted
Site Admin
Posts: 5623
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: Praxis Verbose

Post by ted » Wed Dec 02, 2009 10:49 am

To eliminate printing of "return codes" just wrap the offending statement(s) inside a pair of curly brackets { like so }. Did that help?

roybens
Posts: 44
Joined: Fri Mar 14, 2008 7:57 am

Re: Praxis Verbose

Post by roybens » Fri Dec 04, 2009 7:14 am

Thanks for the fast reply
the problematic line is
minchisq = fit_praxis(NP, "pfunk", &prax_par[0])
when i put the statement in brackets
{minchisq = fit_praxis(NP, "pfunk", &prax_par[0])}
i still get chunks of praxis output like this:

After 73 linear searches, the function has been evaluated 259 times.
The smallest value found is f(x) = 1e+07
log (f(x)) - 0 = 7
x is:41.6812
x is:38.1336
x is:0.766224
x is:1.00488
x is:0.545905
x is:0.480345
After 36 linear searches, the function has been evaluated 92 times.
The smallest value found is f(x) = 5215.98
log (f(x)) - 0 = 3.71734
x is:47.5451
x is:12.3359
x is:-0.031245
x is:0.584925
x is:0.000124178
x is:0.0185188

Any Idea how to get rid of it?
Thanks
Roy

ted
Site Admin
Posts: 5623
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: Praxis Verbose

Post by ted » Fri Dec 04, 2009 10:37 am

Call attr_praxis() with the desired value for printmode before you invoke fit_praxis(). Read about it in the online Programmer's Reference at http://www.neuron.yale.edu/neuron/

roybens
Posts: 44
Joined: Fri Mar 14, 2008 7:57 am

Re: Praxis Verbose

Post by roybens » Fri Dec 04, 2009 11:02 am

I use:

attr_praxis(Tolorance, 0.1,0)
minchisq=0
NUM_ITER = 0
{minchisq = fit_praxis(NP, "pfunk", &prax_par[0])}
PC.post("DONE",iter,NUM_ITER,minchisq)

to call praxis but still get lots of verbose tried it on several versions of neuron and still the printmode parameter does not matter to the amount of verbose NEURON output

hines
Site Admin
Posts: 1580
Joined: Wed May 18, 2005 3:32 pm

Re: Praxis Verbose

Post by hines » Fri Dec 04, 2009 12:32 pm

There was an error in handling the printmode arg of attr_praxis which is fixed in
http://www.neuron.yale.edu/hg/neuron/nr ... 56b39fc92a
(change 'int' to 'long int' in three places).
If you are using the mulrunfitter and want complete quiet, you will need to also
edit nrn/share/lib/hoc/mulfit/fitparm.hoc and comment out the printf statement
in 'proc after_quad()...'
(line 46)

Post Reply