updates to tutorial notebook

Create issue
Issue #2247 closed
Roland Haas created an issue

Please see https://github.com/nds-org/jupyter-et/pull/11

In broad strokes this would change:

  • use %%bash for whole cell instead of ! once per line making copy&paste easier
  • build using -j 2 once more though one needs to check if one does not run out of memory
  • use create-submit rather than create-run
  • include enough data in the plot to provide a preview of the expected plot so that users have a way to check if their run obtained the "correct" results.

This may see further updates as @HelviWitek updates the tutorial partfile to be (again) closer to static_tov.par

Comments (10)

  1. Steven R. Brandt

    When I first made the tutorial, I intentionally used both create-submit and create-run in order to teach that simfactory could either run immediately or in the background.

  2. Roland Haas reporter

    Yes, it means that you can copy multiple bash commands out of the cell into eg a terminal. And as an added bonus even when (incorrectly) copying the %bash line as well then the shells just responds with:

    $ %%bash
    bash: fg: %%bash: no such job

    and continues with the next line (unless, of course, someone has a background process running in that terminal in which case I assume they are advanced Unix users and know what to not copy).

  3. Roland Haas reporter

    In the original tutorial create-run was used for the tov simulation which seems like a bad thing to showcase since create-run cannot be used on clusters. Given choice I would rather shows only a small number of things in this tutorial rather than showcase the full breadth of simfactory. Naturally exactly what one considers the bare minimum to show is up for discussion (and I did add a bit of complex code to get the output directory).

    create-run is useful on ones’s laptop / workstation and I agree with you that showing it there is useful as it given a somewhat easier way of testing things (though I personally very very rarely use it).

    So I would not want to use create-run for the tov simulation for this reason and also b/c it does not leave behind a log file to inspect later. If we want to showcase it (which I am not really averse) I would much prefer if this was done in the "Hello, Cactus!" example rather than the tov example. It would even be ok if the hello world example used no mpirun at all just to show that cactus_sim is just a regular code (though this does, again, not work on most clusters).

  4. Log in to comment