Details
-
Type:
Story
-
Status: Invalid
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: meas_deblender
-
Labels:
-
Story Points:1
-
Team:Data Release Production
Description
Seeing a future warning emanating from meas_deblender when running SourceDeblendTask:
/opt/lsst/software/stack/stack/miniconda3-4.3.21-10a4fa6/Linux64/meas_deblender/14.0-1-g8b7e855+32/python/lsst/meas/deblender/plugins.py:455: FutureWarning: `rcond` parameter will change to the default of machine precision times ``max(M, N)`` where M and N are the input matrix dimensions. |
To use the future default and silence this warning we advise to pass `rcond=None`, to keep using the old, explicitly pass `rcond=-1`. |
X1, r1, rank1, s1 = np.linalg.lstsq(Aw[:, :NT1], bw)
|
Harmless for now, however our demo/tutorials trip it numerous times so it would be great if we could fix it.
Fred Moolekamp Jim told me to assign it to you, don't hate me.
Attachments
Issue Links
Activity
I suggest setting the value to -1 then.
@rhl suggested on slack that we leave this ticket open until we understand why this is being triggered in the first place, as opposed to opening a new ticket to diagnose the problem.
Fred Moolekamp
added a comment - @rhl suggested on slack that we leave this ticket open until we understand why this is being triggered in the first place, as opposed to opening a new ticket to diagnose the problem. We are still getting these warnings (they come out of lsst_dm_stack_demo as well).
I fixed all the ones I saw in meas_deblender on DM-14699. If we're still seeing them in the stack demo, we need new information on where to look for them.
My apologies. I've just checked and it turns out I was using meas_deblender at the commit just before the DM-14699 merge. Sounds like this ticket can be closed.
At the suggestion of Robert Lupton we should confirm that this is the expected behavior. The code in question is attempting to fit a child's flux to a multiple of the PSF. It appears that this can fail if the model doesn't match the PSF. The value of `rcond` determines what to do with singular values, which appears to have changed and has triggered this warning, which is itself in a try except block.
I would recommend not spending too much (any) time on this. This code is not used at all in the new deblender and the time it would take to sufficiently test and understand this behavior does not seem like a good use of time.