Learning Forum

[schinta3]

Hello,

This may be a basic question. Is there a simple way to automatically save the data of all monitors after a simulation finishes?

I know this is possible through scripting to run through the simulation and get all monitor data, or put it into the model>analysis group., but is there a way to make this automatic during a sweep? I.e. auto-save every file to matlab?

[Taylor Robertson]

I think the best way to do this would be to add a script to the ::model analysis tab that collects and save the results you want. If you then create any result here, and make this a result that the sweep collects; then after the sweep simulations have finished it will trigger the analysis script to run, which will save the monitor results to a .mat file

[schinta3]

I see, so this is not possible without the GUI? I.e. if I run the sweep manually by 'saving to files' on a cluster and running them individually?

[Taylor Robertson]

Of course you can save to files and then submit these as batch jobs on the cluster but you would still need the GUI to save the results, and possibly even generate results if they are from a structure group for example. What I would recommend is saving to files, and then submitting the jobs to cluster. Once the simulations have run then you could load from files. If you include the model analysis after you loaded from files then the results would be saved as a set of .mat files.

[schinta3]

Hello
One more question. I am using the fdtd-solutions command line option with Xwindows forwarding on an HPC node. However, the window closes and the .lsf script stops running after ~6min. The error is attached. No output file for the error/exit code from FDTD is given. Is this from an error thrown in Lumerical
or the cluster set up.

[Taylor Robertson]

I've never seen that error before, but it seems like it is related X server node? Running an lsf file requires the CAD, which is graphics application. I would check with your IT department that you can launch a graphics instance on these compute nodes. It is usually trivial to launch engine jobs on these nodes, but there are issues with running lsf without display capabillities. Maybe you could run the lsf locally and submit the jobs to the HPC using the job manager.
https://support.lumerical.com/hc/en-us/articles/360042785654-Running-CAD-jobs-on-headless-compute-instances-Beta-