I am not sure if this change is documented or is an intuitive change. However, it would be nice to have it on the comparison page.

Already working on it :-)

Cheers,
Jochen!



On Tue, Nov 30, 2021 at 9:36 AM Jochen Martin Eppler <j.eppler@fz-juelich.de> wrote:

Dear Maryada,

[Re-sending with properly rendered Markdown. Sorry for the noise]


I re-tested my code and realized that I didn't set overwrite_files to False properly, so you are right, I received the error.

Good.

However, since I am importing my code from NEST 2.xx version to v3.1 and all my previous implementation also have similar two-phase simulations, this was not the case earlier. To my understanding, unless you call ResetKernel, it is a single simulation and should not overwrite the current file as is the case when recording in memory.

The behavior has indeed been as you describe with NEST 2.x. However, it was changed with NEST 3.0.

In the 2.x series of NEST, all recording devices had a flag close_after_simulate, which defaulted to False. As this flag resulted in a high bookkeeping complexity internally and was confusing to many users, it has been removed in favor of the clearer semantics that each call to Simulate() produces its own file.

I ran your script (with syntax changes) on my laptop which only has nest 2.20 version and the attached data file contains recording for the whole run and not just the last subpart of simulation. I would understand if the functionality is changed in the latest version, however, I was wondering if there is a possible work-around. I surely cannot manually store all the data given that I also record from a multimeter and these are memory-consuming simulations.

If you want to run repetitive simulations and get all your data (either in separate files or a single file) with NEST 3.x, you have several possibilities:

  1. You change the kernel property data_path inbetween calls to Simulate(). This would let NEST put the data for all recorders in different directories. You can use os.mkdir to create the directories from within your simulation script:

    import nest, os
    # Create nodes
    # Connect nodes
    os.mkdir("path_for_sim1")
    nest.data_path = "path_for_sim1"
    nest.Simulate(1000)
    os.mkdir("path_for_sim2")
    nest.data_path = "path_for_sim2"
    nest.Simulate(2000)
    
  2. You can move your files to somewhere safe using os.rename(), instead of setting data_path.

  3. You can set a different data_prefix between calls to Simulate() (i.e., nest.data_prefix = "sim1_", nest.data_prefix = "sim2_", …). That way you wouldn’t have to create directories, but your data files would instead be prefixed as to your choice.

  4. You can use Prepare(), Run(), Cleanup() instead of Simulate() as explained in the Guide on running simulations. For each call to Run(), the recorders would actually append to the files, as those are only closed in Cleanup() (and re-opened in Prepare()). In fact, Simulate() is actually just a wrapper around the three aforementioned functions.

Cheers,
Jochen!



On Mon, Nov 29, 2021 at 10:24 PM Jochen Martin Eppler <j.eppler@fz-juelich.de> wrote:

Dear Maryada,

I have just played a bit with your example (see attachment for a slightly extended and reformatted version) and as far as I can see, NEST behaves exactly as expected (and documented).

Please let me explain:

  • When using the recording backend ascii with the kernel attribute overwrite_files set to True, you have to move files away at the location in your script saying do something, otherwise the data files get overwritten by each new call to Simulate() and you consequently get to see only the results from the last run (which is what you report).

  • If you set overwrite_files to False, and did not move the file away, you will get an error upon repeated calls to Simulate(), as the file to be created by the backend already exists.

  • When using the recording backend memory, overwrite_files is not even considered, as that does not actually write files.

  • Data just accumulates in the recording backend memory until it is explicitly cleared by running Input_SD.n_events = 0.

See the linked backend documentation pages for more information.

I hope this helps to clear up things.

Cheers,
Jochen!

On 29.11.21 18:32, Maryada Maryada wrote:

Dear NEST users,

I am trying to debug my network and I found that my spike_recorder with ascii option isn't working as expected.
I am simulating a network for a few milliseconds (say 1000. ms) and then do some structure changes and then re-run for another phase of a few ms (say 4000. ms)
In the compact version, I am trying to see how this sample code would work.
import nest
nest.ResetKernel()
nest.SetKernelStatus({
                     "overwrite_files": True,
                      "data_path": './Debug-Log',
                      "data_prefix": 'DEBUG-SPIKE-REC',
                   })
noise_device = nest.Create("poisson_generator",params={'start': 0.,'rate' : 100.})
sd_params = {
            "record_to": 'ascii',
            "label": "Input_spike_recorder"
            }
Input_SD = nest.Create("spike_recorder",1, params=sd_params)
nest.Connect(noise_device,Input_SD)
nest.Simulate(100.)
# do something here
nest.Simulate(100.)

Issue: I only get the data recorded from 100th millisecond but for memory, I get all the data from 0-200 ms.
This behavior is consistent for both overwrite_files as True and False.
In the previous version (v2.20) I also used the same approach and had no issues. Is there a newer way to do such a simulation paradigm?
I didn't find anything on this on the comparison page of  NESTv2.x and NESTv3.x

Am I missing something? It's really important for my work, I hope I could find a solution.
--
Thanks and Regards

Maryada


_______________________________________________
NEST Users mailing list -- users@nest-simulator.org
To unsubscribe send an email to users-leave@nest-simulator.org
-- 
Dr. Jochen Martin Eppler
Phone: +49(2461)61-96653
----------------------------------
Simulation Laboratory Neuroscience
Jülich Supercomputing Centre
Institute for Advanced Simulation


------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------
Forschungszentrum Juelich GmbH
52425 Juelich
Sitz der Gesellschaft: Juelich
Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
Vorsitzender des Aufsichtsrats: MinDir Volker Rieke
Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender),
Karsten Beneke (stellv. Vorsitzender), Prof. Dr. Astrid Lambrecht,
Prof. Dr. Frauke Melchior
------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------

_______________________________________________
NEST Users mailing list -- users@nest-simulator.org
To unsubscribe send an email to users-leave@nest-simulator.org


--
Thanks and Regards

Maryada


_______________________________________________
NEST Users mailing list -- users@nest-simulator.org
To unsubscribe send an email to users-leave@nest-simulator.org
-- 
Dr. Jochen Martin Eppler
Phone: +49(2461)61-96653
----------------------------------
Simulation Laboratory Neuroscience
Jülich Supercomputing Centre
Institute for Advanced Simulation
_______________________________________________
NEST Users mailing list -- users@nest-simulator.org
To unsubscribe send an email to users-leave@nest-simulator.org


--
Thanks and Regards

Maryada


_______________________________________________
NEST Users mailing list -- users@nest-simulator.org
To unsubscribe send an email to users-leave@nest-simulator.org
-- 
Dr. Jochen Martin Eppler
Phone: +49(2461)61-96653
----------------------------------
Simulation Laboratory Neuroscience
Jülich Supercomputing Centre
Institute for Advanced Simulation