Hi.
relax does take much longer than both Dasha and Modelfree. The reason
for this is that the optimisation precision in relax is many, many
orders of magnitude higher than both Dasha and Modelfree. This makes
a significant difference to the final results.
Has relax now returned results? On a 160 residue protein, the
completion of the new model-free analysis protocol encoded in
'full_analysis.py' took just over one week to complete. This was on a
dual processor, dual core Intel Xeon 2.8 GHz machine. And that
included 14 rounds of optimisation for the spherical, oblate
spheroidal, prolate spheroidal, and ellipsoidal diffusion models.
That would be ~14 hours per iteration per diffusion model. Do you
mind me asking what the CPU speeds of the Pelican, Owl, and Pigeon is?
As you have 220 residues (the effect of the number of NMR relaxation
data sets on the computation speed would probably be negligible), it
may take longer than a week to complete the calculations (depending on
your CPU speeds).
Also did relax create a directory called 'm8'? Each additional
parameter in the model increases the CPU time required by an order of
magnitude because it adds one additional dimension to the chi-squared
space. For the grid search, it will be almost exactly 1 order of
magnitude; the optimisation afterwards would be much less than 1 order
of magnitude. The time required for model 'm8' may be close to or
above that required for the sum of all other models.
Hopefully the calculations of the first round are now complete.
Otherwise, there may be an issue somewhere slowing things down. If
that is the case, then we can hunt down the issue and fix it.
Bye,
Edward
On 10/13/06, Sebastien Morin <sebastien.morin.1@xxxxxxxxx> wrote:
Hi !
I am trying the script full_analysis.py with my data.
I work with data at three different magnetic fields for 220 residues on
a 271 residue protein.
I successfully ran the script with the local_tm model, and then with all
other diffusion models (sphere, prolate, oblate and ellipsoid) as what
concerns the initial run.
However, the round 1 runs seem to be stuck (or maybe am I too
impatient)... Let me explain with some benchmarks.
Model Round Machine Time
========== ===== =========== =======
local_tm --- Pelican 15h00
-----------------------------------------------
sphere init Pelican 0h10
1 Pelican > 30h00
prolate init Hibou 0h20
1 Hibou > 30h00
oblate init Hibou 0h20
1 Hibou > 30h00
ellipsoid init Pigeon 1h00
1 > 30h00
-----------------------------------------------
final init
Also, I find strange the fact that all sub-directories (m0, m1, m2, m3,
m4, m5, m6, m7, m8, m9 and aic) in those pseudo stuck runs were created
with their bz2 archives more than 24h00 ago...
Moreover, it's been more than 6h00 since nothing hasn't been put in the
log files (which are redirected screen text).
So...
1. Is it normal that these runs take so long to stop ?
2. How can I know that the runs aren't stuck ?
Thank you for your precious help !
Sébastien :)
_______________________________________________
relax (http://nmr-relax.com)
This is the relax-users mailing list
relax-users@xxxxxxx
To unsubscribe from this list, get a password
reminder, or change your subscription options,
visit the list information page at
https://mail.gna.org/listinfo/relax-users
Re: full_analysis.py