Allow large blends. Add error fields #218
Conversation
This avoids getting spammed by warnings in new stacks
also some vis
See if this fixes error in CI AttributeError: type object 'lsst.afw.table._table.CoordKey' has no attribute 'addErrorFields'
|
It seems detection may have changed, either due to stack or something we did. Previously objects were detected in zero weight (inf variance) regions but not now. This test |
The defaults changed and broke things, so set everything Add test case where BRIGHT is set and inf var
This fix was put into ngmix back in Oct 2023
we added a new sum for AM that needs to be skipped for this
|
The |
|
summary of additional changes
|
|
@beckermr can you have a look at this one? |
|
Shall we merge @esheldon? |
| keys = config.toDict().keys() | ||
|
|
||
| for key in keys: | ||
| assert key in expected_keys, f'found unexpected config key {key}' |
There was a problem hiding this comment.
I am not sure what this is meant to achieve. It will fail the moment we add a new field to this config, but won't if we remove a config. DM guarantees to deprecate a field before removing and not remove a field until the next official release. DM makes no such guarantee when it comes to adding new fields.
There was a problem hiding this comment.
The point is that if something changes, we need to understand what that change is and ensure it works. We shouldn't simply let things float.
|
I saw a resolved conversation after I posted this comment and understood
the motivation.
…On Wed, Aug 28, 2024, 1:01 PM Matthew R. Becker ***@***.***> wrote:
***@***.**** commented on this pull request.
------------------------------
In metadetect/lsst/tests/test_lsst_dm_configs.py
<#218 (comment)>:
> + 'tempLocalBackground',
+ 'doTempLocalBackground',
+ 'tempWideBackground',
+ 'doTempWideBackground',
+ 'nPeaksMaxSimple',
+ 'nSigmaForKernel',
+ 'statsMask',
+ 'excludeMaskPlanes'
+ ]
+
+ config = SourceDetectionConfig()
+
+ keys = config.toDict().keys()
+
+ for key in keys:
+ assert key in expected_keys, f'found unexpected config key {key}'
The point is that if something changes, we need to understand what that
change is and ensure it works. We shouldn't simply let things float.
—
Reply to this email directly, view it on GitHub
<#218 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABG56HRODSPWNO5VHD6DNNDZTX643AVCNFSM6AAAAABJ76S5U2VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDENRWHAZDKNJZHA>
.
You are receiving this because you commented.Message ID:
***@***.***>
|
This deblending change is important. It should result in a lot more detections in blended regions, and it potentially requires a full retesting.
We might be happy retesting in "no warp" mode, direct to coadd simulations, which would speed it up by a factor of perhaps 2.5