Fwd: Run Modeller processes on multiple cpus
Hi,
Greetings....
I am running mod9v11 on Intel Xeon with 12 cores processor. I have found that modeller is utilizing only single processor. Even though there are few older posts regarding the execution of modeller processors on multiple processors, I am unable to implement them. Sorry, I am newbie to python.
It will be a great help if you could help me with a script or procedure to set up the running of modeller processes on multiple processors. This will save my lifetime. Please do the needful.
Thank you
Regards,
Lalith
On 4/19/13 1:31 AM, Lalith Kumar wrote: > I am running mod9v11 on Intel Xeon with 12 cores processor. I have found > that modeller is utilizing only single processor. Even though there are > few older posts regarding the execution of modeller processors on > multiple processors, I am unable to implement them. Sorry, I am newbie > to python.
Since you don't say what you've tried, or what problems you ran into, it's difficult to help you. Modeller doesn't use all the processors by default. To take advantage of multiple cores, you can either split your job manually (as Oscar suggests) or have Modeller do it for you automatically. To do the latter, you need to set up a parallel job object and then call a.use_parallel_job(). How you create the parallel job object depends on how you want to parallelize your job (e.g. whether you want to use a batch scheduling system). To just use 12 cores on the local machine, something like the following should work:
a = automodel(...) j = job(host='localhost') for i in range(12): j.append(local_slave()) a.use_parallel_job(j)
Ben Webb, Modeller Caretaker
Dear Ben and Oscar,
Thank you for the suggestions. Sorry if my query is not clear. To explain, I am trying to do loop modeling and at some situation I have to do cooperatively for about 12 loops. So, here I need to implement the multiple processing in modeller. I am here attaching the script (for single loop) that I have prepared to do on my 12 core processor. But, unfortunately I could not execute this because of the python script error, which I could not trace. From this point, I would like to get suggestion and help from your side. Please do the needful correction to my scripts and let me know where did I went wrong.
Thank you a Lot
On Fri, Apr 19, 2013 at 9:28 PM, Modeller Caretaker < modeller-care@salilab.org> wrote:
> On 4/19/13 1:31 AM, Lalith Kumar wrote: > >> I am running mod9v11 on Intel Xeon with 12 cores processor. I have found >> that modeller is utilizing only single processor. Even though there are >> few older posts regarding the execution of modeller processors on >> multiple processors, I am unable to implement them. Sorry, I am newbie >> to python. >> > > Since you don't say what you've tried, or what problems you ran into, it's > difficult to help you. Modeller doesn't use all the processors by default. > To take advantage of multiple cores, you can either split your job manually > (as Oscar suggests) or have Modeller do it for you automatically. To do the > latter, you need to set up a parallel job object and then call > a.use_parallel_job(). How you create the parallel job object depends on how > you want to parallelize your job (e.g. whether you want to use a batch > scheduling system). To just use 12 cores on the local machine, something > like the following should work: > > a = automodel(...) > j = job(host='localhost') > for i in range(12): > j.append(local_slave()) > a.use_parallel_job(j) > > Ben Webb, Modeller Caretaker > -- > modeller-care@salilab.org http://www.salilab.org/**modeller/http://www.salilab.org/modeller/ > Modeller mail list: http://salilab.org/mailman/**listinfo/modeller_usagehttp://salilab.org/mailman/listinfo/modeller_usage >
p.s: For corrections such as - *Leave out the argument to the job() constructor *I am getting the error "name 'job' is not defined" and I am running with the python command instead of mod. (i.e., python2.6 looprefine.py > looprefine.log)
On Sat, Apr 20, 2013 at 2:36 PM, Lalith Kumar lalithkumar07@gmail.comwrote:
> Dear Ben and Oscar, > > Thank you for the suggestions. Sorry if my query is not clear. To explain, > I am trying to do loop modeling and at some situation I have to do > cooperatively for about 12 loops. So, here I need to implement the multiple > processing in modeller. I am here attaching the script (for single loop) > that I have prepared to do on my 12 core processor. But, unfortunately I > could not execute this because of the python script error, which I could > not trace. From this point, I would like to get suggestion and help from > your side. Please do the needful correction to my scripts and let me know > where did I went wrong. > > Thank you a Lot > > > On Fri, Apr 19, 2013 at 9:28 PM, Modeller Caretaker < > modeller-care@salilab.org> wrote: > >> On 4/19/13 1:31 AM, Lalith Kumar wrote: >> >>> I am running mod9v11 on Intel Xeon with 12 cores processor. I have found >>> that modeller is utilizing only single processor. Even though there are >>> few older posts regarding the execution of modeller processors on >>> multiple processors, I am unable to implement them. Sorry, I am newbie >>> to python. >>> >> >> Since you don't say what you've tried, or what problems you ran into, >> it's difficult to help you. Modeller doesn't use all the processors by >> default. To take advantage of multiple cores, you can either split your job >> manually (as Oscar suggests) or have Modeller do it for you automatically. >> To do the latter, you need to set up a parallel job object and then call >> a.use_parallel_job(). How you create the parallel job object depends on how >> you want to parallelize your job (e.g. whether you want to use a batch >> scheduling system). To just use 12 cores on the local machine, something >> like the following should work: >> >> a = automodel(...) >> j = job(host='localhost') >> for i in range(12): >> j.append(local_slave()) >> a.use_parallel_job(j) >> >> Ben Webb, Modeller Caretaker >> -- >> modeller-care@salilab.org http://www.salilab.org/**modeller/http://www.salilab.org/modeller/ >> Modeller mail list: http://salilab.org/mailman/**listinfo/modeller_usagehttp://salilab.org/mailman/listinfo/modeller_usage >> > > > > -- > Lalith Kumar K > Research Associate, > CADD Division, > Dr. Reddy's Institute of Life Sciences (Formerly known as Institute of > Life Sciences) > University of Hyderabad Campus > Gachibowli, Hyderabad 500046, India > Tel: +91 9502074040 >
On 4/20/13 2:06 AM, Lalith Kumar wrote: > Thank you for the suggestions. Sorry if my query is not clear. To > explain, I am trying to do loop modeling and at some situation I have to > do cooperatively for about 12 loops.
Right, but you still haven't told me what the error is.
Looking at your script, you probably need from modeller.parallel import * at the start. And you probably don't need to set the modeller_path argument in the job constructor.
Ben Webb, Modeller Caretaker
participants (3)
-
Lalith Kumar -
Modeller Caretaker -
Oscar Conchillo Solé