Skip to first unread message
Prouteau Manoel
May 10, 2016, 5:43:56 AM5/10/16
to emspring
Dear Carsten,
I recently started to use again Spring (it took me time but I have a good set of cryo-EM micrographes). Nevertheless, I have few basic issues about Spring.
We have recently Downloaded Spring on our university server (http://baobabmaster.unige.ch/enduser/src/enduser/) and I am facing problems with the SegmentClass command. The server requires to launch jobs via sbatch; so the last sbatch.sh I used is attached. I also attach the parameterfile.txt, the report.log and the logfile. I have few hypothesis about the problem that I am facing:
1- I am not at all expert of mpi use but I am wondering if Spring is correctly matching with the server structure. Indeed, it appears that I can not spread the job on more than one node. Indeed, if I apply in the sbatch.sh #SBATCH --ntask=n and using openmpi/gcc it launches n times the SegmentClass instead of once. At the opposite when I used one node (sbatch.sh #SBATCH --ntask=1, cpu=16 without loading openmpi/gcc) the SegementClass started correctly but rapidly stopped (cf report.log and logfile). So is the problem linked to the mpi stucture compatibility with the server? Can we spread the Spring job over several nodes? If yes what are the requirements from the server (in case the setting is complex could I put you in contact with the manager of the server?)?
2- The alternative and not exclusive explanation would be about that the calculation power needed. Indeed I found an old post with a similar issue (Venkata 14/01/23) where you mentioned decreasing binning factor, boxes sizes,... I guess that this is to reduce the volume of data used during the procedure. I already decreased the binning to 3 (cf. parameterfile.txt) without success. I have to say that the helices I am working on are quite big (estimated width: 580 A, estimated Helix step: 450 A). Taking in account these sizes, what would be, for you, the best segmentation/classification strategies?
I hope my questions make sense and that I am not bothering you too much.
Waiting for your help, have a great day.
Manoël
I recently started to use again Spring (it took me time but I have a good set of cryo-EM micrographes). Nevertheless, I have few basic issues about Spring.
We have recently Downloaded Spring on our university server (http://baobabmaster.unige.ch/enduser/src/enduser/) and I am facing problems with the SegmentClass command. The server requires to launch jobs via sbatch; so the last sbatch.sh I used is attached. I also attach the parameterfile.txt, the report.log and the logfile. I have few hypothesis about the problem that I am facing:
1- I am not at all expert of mpi use but I am wondering if Spring is correctly matching with the server structure. Indeed, it appears that I can not spread the job on more than one node. Indeed, if I apply in the sbatch.sh #SBATCH --ntask=n and using openmpi/gcc it launches n times the SegmentClass instead of once. At the opposite when I used one node (sbatch.sh #SBATCH --ntask=1, cpu=16 without loading openmpi/gcc) the SegementClass started correctly but rapidly stopped (cf report.log and logfile). So is the problem linked to the mpi stucture compatibility with the server? Can we spread the Spring job over several nodes? If yes what are the requirements from the server (in case the setting is complex could I put you in contact with the manager of the server?)?
2- The alternative and not exclusive explanation would be about that the calculation power needed. Indeed I found an old post with a similar issue (Venkata 14/01/23) where you mentioned decreasing binning factor, boxes sizes,... I guess that this is to reduce the volume of data used during the procedure. I already decreased the binning to 3 (cf. parameterfile.txt) without success. I have to say that the helices I am working on are quite big (estimated width: 580 A, estimated Helix step: 450 A). Taking in account these sizes, what would be, for you, the best segmentation/classification strategies?
I hope my questions make sense and that I am not bothering you too much.
Waiting for your help, have a great day.
Manoël
Prouteau Manoel
May 13, 2016, 6:15:51 AM5/13/16
to emspring
Dear Carsten,
I do not know if you have seen my previous post. I actually found a transient solution. Indeed it appears that the compilation of Spring has some problem of compatibility with our server here. I could go through Spring2D.
Hope everything will be fine now.
Best regards,
Manoël Prouteau
I do not know if you have seen my previous post. I actually found a transient solution. Indeed it appears that the compilation of Spring has some problem of compatibility with our server here. I could go through Spring2D.
Hope everything will be fine now.
Best regards,
Manoël Prouteau
Reply all
Reply to author
Forward
0 new messages