Dynamite Core Team Meeting 19th April

[boeckera]

Hola @dynamics-of-stellar-systems/dynamite-maintainers,

at this week's meeting at 1pm (onsite: Glenn's office, online: dynamite zoom link) we will mainly talk about the issue of storage space on VSC and possible deletion of some DYNAMITE files that are currently saved automatically.

Please, if you have any other topics you want to bring up respond below.
To indicate your participation you can react with an appropriate emoji of your choice.

Looking forward to seeing you all!

Un saludo,
Alina

[maindlt]

Hi there, does anyone have time to test PR #365 (about labels in the chi2 plot, would be a quick win 😉) and/or have quick look at PR #379 (about citing Decomposition in the docs)...? Cheers and see you Friday, Thomas

[boeckera]

Hi Thomas,

could you kindly describe a quick instruction again how to test this? I installed dynamite on my MAC, but now what do I do haha
We can also go through it together on Friday :)

Thank you!

[maindlt]

Hi Alina,

many thanks for taking the time to test! :-)

PR #365 is all about the $\chi^2$ plot which is stored in plots/kinchi2_plot.png in the output directory. More specifically, the PR fixes the labels of the x and y axes and the colorbar and also makes sure the colorbar does not overlap with the actual plot.

We can of course do this together on Friday, but if you are interested please have a look at my ideas for testing below ;-)

Recommended testing:

• Check out the pull request's branch chi2_plot_fixes. If DYNAMITE was installed already, the legacy Fortran does not need to be recompiled for this PR, but a python setup.py install --user is still necessary to update all the Python files.
• Let's use the test script test_nnls.py in the dev_tests/ folder for testing. This script reads the config file user_test_config_ml.yaml in the same directory.
• Running the script will create an empty $\chi^2$ plot because all parameters but one (ml) are fixed. To create a minimal $\chi^2$ plot, one additional parameter needs to be set to fixed: False in the configuration file. Before, this created a plot where the colorbar overlapped the x-axis. This should not be the case any more here. I used fixed: False for the black hole m parameter, perhaps you can use another one, run the script with this new configuration, and verify that the colorbar no longer overlaps and has a meaningful caption.
• Finally, let's create a more complex plot and verify that all the captions make sense. For this, I increased the maximum number of models to n_max_mods : 6 in the parameter_space_settings of the config file and set a few more parameters to fixed: False (I chose the dark halo c and the stars q, but any parameter that has some entries in its par_generator_settings: qualifies). Running the script again takes a bit longer (about 2.25 min on my 2020 Mac) and should result in a more complex plots/kinchi2_plot.png with meaningful labels on the axes and the colorbar.
• Done :-)

Cheers,
Thomas