Right now, the multiprocessor code is in a development branch. But you can run each global diffusion model one by one on different machines.
Regards, Edward On Wed, Nov 5, 2008 at 2:37 PM, Tyler Reddy <[EMAIL PROTECTED]> wrote: > Interesting. I'm guessing it's alright to optimize the different models at > the > same time on different CPUs since models II-V can be done in any order. Is > there a way to specify use of another CPU on a parallel processor machine, > or > should I just run relax on other machines? > > Tyler > > > Quoting Edward d'Auvergne <[EMAIL PROTECTED]>: > >> Hi, >> >> 14 rounds of that length is quite normal. See Seb's post at >> https://mail.gna.org/public/relax-users/2006-11/msg00015.html for >> typical details. Actually, all the posts in the thread starting at >> https://mail.gna.org/public/relax-users/2006-11/msg00014.html could be >> useful. You can make this as fast as Art Palmer's Modelfree program >> >> (http://cpmcnet.columbia.edu/dept/gsas/biochem/labs/palmer/software/modelfree.html), >> but I can also modify Modelfree to be almost as slow as relax. relax >> will always be a little slower than Modelfree as the model-free code >> is in numpy which is slower than Modelfree's FORTRAN. Converting this >> relax code will speed relax up a bit. But, anyway, the major timing >> difference you see if running the 2 programs on default settings is >> due to an absolutely massive difference in optimization precision. >> For details please see d'Auvergne, E. J. and Gooley, P. R. 2008a >> (http://www.nmr-relax.com/refs.html). >> >> Regards, >> >> Edward >> >> >> On Fri, Oct 31, 2008 at 5:47 PM, Tyler Reddy <[EMAIL PROTECTED]> wrote: >>> >>> I'm on the 14th round of 'prolate' optimization at several hours per >>> round. I >>> guess I'll have to try and get errors a different way as we discussed for >>> this >>> to be reasonable in relax. >>> >>> Tyler >>> >>> >>> >>> Quoting Edward d'Auvergne <[EMAIL PROTECTED]>: >>> >>>> Hi, >>>> >>>> You'll need to update using: >>>> >>>> $ svn up >>>> >>>> then install as root/administrator by typing (assuming python is >>>> installed as root): >>>> >>>> $ python setup.py install >>>> >>>> in the minfx base directory. Then the crash shouldn't happen. If it >>>> does still have problems, it would be worth reporting it to the minfx >>>> mailing lists (https://gna.org/mail/?group=minfx). A bug report for >>>> minfx can be submitted at >>>> https://gna.org/bugs/?func=additem&group=minfx. >>>> >>>> As for this taking a long time to optimise, that is because of the >>>> convoluted, very tight optimisation space. You can decrease the >>>> max_iter value from 1e7 in the minimise() user function at the start >>>> to to maybe 1e5, but this should not be done at the end of the >>>> analysis. I would say though that all these problems will go away >>>> with a better error analysis in the R2 curve-fitting. >>>> >>>> Regards, >>>> >>>> Edward >>>> >>>> >>>> >>>> On Mon, Oct 27, 2008 at 12:48 AM, Tyler Reddy <[EMAIL PROTECTED]> wrote: >>>>> >>>>> I used subversion to update minfx, but I'm still having the problem >>>>> (unless I >>>>> need to delete the old minfx, or does the nested update automatically >>>>> 'take >>>>> over' in this case?). >>>>> >>>>> The calculation certainly goes a lot longer now than it did before, but >>>>> at k=312 >>>>> it crashes out with the same traceback. Another possibility is that >>>>> I have to >>>>> delete all the local_tm and sphere folders and just start from scratch >>>>> with the >>>>> new minfx, though I'm not sure that makes sense. >>>>> >>>>> Tyler >>>>> >>>>> >>>>> Quoting Edward d'Auvergne <[EMAIL PROTECTED]>: >>>>> >>>>>> Right, I've fixed the bug in the minfx repository. It will take me a >>>>>> while to release minfx 1.0.2 but using subversion you can get the >>>>>> changes immediately. Note that as this space is so warped, >>>>>> optimisation of this model will take a long, long time! >>>>>> >>>>>> Regards, >>>>>> >>>>>> Edward >>>>>> >>>>>> >>>>>> On Fri, Oct 24, 2008 at 10:40 PM, Edward d Auvergne >>>>>> <[EMAIL PROTECTED]> wrote: >>>>>>> >>>>>>> Update of bug #12499 (project relax): >>>>>>> >>>>>>> Status: Confirmed => Fixed >>>>>>> Open/Closed: Open => Closed >>>>>>> >>>>>>> _______________________________________________________ >>>>>>> >>>>>>> Follow-up Comment #5: >>>>>>> >>>>>>> This problem has been fixed in the minfx project >>>>>>> (https://gna.org/projects/minfx/), in the bug report >>>>>>> https://gna.org/bugs/?12501. >>>>>>> >>>>>>> _______________________________________________________ >>>>>>> >>>>>>> Reply to this item at: >>>>>>> >>>>>>> <http://gna.org/bugs/?12499> >>>>>>> >>>>>>> _______________________________________________ >>>>>>> Message sent via/by Gna! >>>>>>> http://gna.org/ >>>>>>> >>>>>>> >>>>>> >>>>> >>>>> >>>>> >>>>> >>>>> _______________________________________________ >>>>> relax (http://nmr-relax.com) >>>>> >>>>> This is the relax-devel mailing list >>>>> [email protected] >>>>> >>>>> 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-devel >>>>> >>>> >>>> _______________________________________________ >>>> relax (http://nmr-relax.com) >>>> >>>> This is the relax-devel mailing list >>>> [email protected] >>>> >>>> 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-devel >>>> >>> >>> >>> >>> >>> _______________________________________________ >>> relax (http://nmr-relax.com) >>> >>> This is the relax-devel mailing list >>> [email protected] >>> >>> 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-devel >>> >> > > > > _______________________________________________ relax (http://nmr-relax.com) This is the relax-devel mailing list [email protected] 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-devel
