Initialization error in docs/tutorials/multiresolution.ipynb

[b-remy]

I was running the multiresolution.ipynb tutorial and encountered an error when initializing the sources as ExtendedSource

# Source initialisation
sources = [
    scarlet.ExtendedSource(model_frame,
                           sky_coord,
                           observations,
                           thresh=0.1,
                          )
    for sky_coord in ra_dec
]

returning

ArithmeticError: Model TabulatedSpectrum, Parameter 'spectrum' is not finite:
[      nan       nan       nan       nan       nan 2.6395272]

Has scarlet initialization procedure changed after this notebook was pushed? Here is a colab notebook to reproduce the error https://colab.research.google.com/drive/1nV-mGIQoCJZbExxPD_usa0evmJIZjOG8?usp=sharing

[pmelchior]

Weird. This is the spectrum initialization of the HSC image going wrong. Can you check if the pixel coordinates associated with the source position in RA/Dec is correct?

[b-remy]

It seems they are in RA/DEC

array([[150.33104781,   2.07003218],
       [150.33022577,   2.0702573 ],
       [150.33102661,   2.07022809],
       [150.33106896,   2.07029557],
       [150.33096789,   2.07025819],
       [150.33104769,   2.0701253 ],
       [150.33005928,   2.07028169],
       [150.32999003,   2.07030221],
       [150.32993349,   2.07048086],
       [150.32991772,   2.07056598],
       [150.33021891,   2.07083053],
       [150.33059818,   2.07098648]])

And the display function is converting these RA/DEC coordinates to HSC pixels to overlap sources to the morphology

I am trying to figure out where the nans come from

[pmelchior]

I think I have found part of the problem. If so, it's related to #282. Can you test if spectra = init.get_pixel_spectrum(sky_coord, observations) returns nans, please.

[b-remy]

This works for each source

[array([4.375723 , 3.706623 , 2.6916485, 3.2171175, 2.4734876],
      dtype=float32), array([1.4452889], dtype=float32)]
[array([8.453968, 8.744746, 8.078425, 7.21597 , 6.92792 ], dtype=float32), array([6.1424046], dtype=float32)]
[array([10.523911 ,  9.074745 ,  6.5678105,  7.0109396,  4.94793  ],
      dtype=float32), array([4.0687194], dtype=float32)]
[array([9.881581 , 8.826522 , 6.136643 , 6.26985  , 4.0761013],
      dtype=float32), array([13.809645], dtype=float32)]
[array([8.960169, 7.722469, 5.366055, 5.421723, 4.023126], dtype=float32), array([3.0977886], dtype=float32)]
[array([7.3961725, 6.314975 , 4.6239867, 5.7644944, 4.153218 ],
      dtype=float32), array([4.3545218], dtype=float32)]
[array([18.904358, 21.548618, 21.980566, 22.801405, 18.784016],
      dtype=float32), array([23.688356], dtype=float32)]
[array([21.445032, 24.591253, 24.96186 , 25.030066, 20.297144],
      dtype=float32), array([16.399578], dtype=float32)]
[array([5.963582 , 6.2947664, 6.730194 , 7.1471443, 6.6866164],
      dtype=float32), array([3.1061878], dtype=float32)]
[array([4.42793  , 4.904378 , 5.2223344, 5.5198197, 5.395245 ],
      dtype=float32), array([2.4887385], dtype=float32)]
[array([1.1586634, 3.9580977, 8.05664  , 9.940866 , 8.854514 ],
      dtype=float32), array([9.527296], dtype=float32)]
[array([0.48858365, 0.92691386, 0.74116206, 0.7110228 , 0.5247147 ],
      dtype=float32), array([0.9085161], dtype=float32)]

[pmelchior]

Check out brach init_fixes. It should work now. I haven't run all tests, but if they work as expected I will create a PR for that branch asap

[b-remy]

Thank you for the fix @pmelchior the tutorial notebook works fine now! I'm closing this issue