This is a relax source code issue which is only triggered by running
the 'model-free.py' script with MC sims then running the 'modsel.py'
script. The 'full_analysis.py' script will not be affected.
As for the computation time of the 'full_analysis.py' script, because
of relax's high precision (its default precision is orders of
magnitude greater than Modelfree4 and Dasha) and the use of the higher
models m7 and m8, it does take a while. For example on a 160 residue
protein using relaxation data collected at two field strengths, it
took just over a week to finish calculations on a dual CPU, dual core
Xeon 2.8 GHz system. This 'full_analysis.py' script implements a
completely new protocol for obtaining the model-free description.
Instead of starting with the diffusion tensor estimate, the first part
of the protocol is the optimisation of the model-free parameters free
of any global parameters. There are significant benefits to this
approach, most of which I have explained in papers that I am in the
process of getting published. These benefits should outweigh the
disadvantage of the computation time. The full protocol is described
in a flow diagram I have added to the relax manual (in the 1.2 and 1.3
lines).
One last thing that is very important - the new model-free
optimisation protocol embodied by the 'full_analysis.py' script
requires relaxation data at a minimum of two field strengths. All of
the problems that it solves depend on this. If single field strength
data is input, garbage will be output :)
Cheers,
Edward
On 11/22/06, Headey, Stephen <S.J.Headey@xxxxxxxxxxxx> wrote:
Is this a data problem or a source problem? I'm also running full_analysis.py
and the second iteration of 'sphere' is still running after 18hrs. Is this
expected? or could this be a related problem?
-----Original Message-----
From: Edward d Auvergne [mailto:NO-REPLY.INVALID-ADDRESS@xxxxxxx]
Sent: Wed 11/22/2006 5:38 AM
To: Edward d Auvergne; Headey, Stephen; relax-devel@xxxxxxx
Cc:
Subject: [bug #7755] modsel.py error with "eliminate"
Update of bug #7755 (project relax):
Status: None => Confirmed
Operating System: GNU/Linux => All systems
_______________________________________________________
Follow-up Comment #1:
The problem here is that the Monte Carlo simulation data structures are not
being properly recreated by the model-free results reading functions. I
tried this using 10 Monte Carlo simultations together with the relaxation
data attached to bug #7787 (https://gna.org/bugs/?7787). After reading the
results files, self.relax.data.sim_number[run] was set to 1300! This is 10
times 130 spins! Hence elements beyond the end of the Monte Carlo simulation
arrays (of length 10) are being accessed.
_______________________________________________________
Reply to this item at:
<http://gna.org/bugs/?7755>
_______________________________________________
Message sent via/by Gna!
http://gna.org/