BlueJeans: https://bluejeans.com/426716450
attending: Eric Bellm Ian Sullivan John Swinbank Meredith Rawls Krzysztof FindeisenUnknown User (cmorrison)
Topics for discussion:
- HiTS reprocessing (Meredith Rawls)
- summarizing recent work on , which spawned ,
- Chris suggests that flags labeled "slot" may be more useful than eg using the pixel flags.
- Ian suggests that DM-16544 is due to using PSF-matched coadds–any time the PSF convolution kernel touches a mask. But that's to make the coadd, and the mask growing is happening in diffim.
- Krzysztof says Yusra recommend convolvedTemplate=True last year for use with PsfMatched coadds. This is the default setting for
ImageDifferenceTask
, so we're currently using it implicitly.
- DES processing (Unknown User (emorganson))
- CI (Krzysztof Findeisen )
- SQuaRE has been pushing forward on pushing ap_verify metrics to SQuASH. But SQuASH assumes some metadata is present that isn't there. Simon Krughoff has promised to fix it in
- DCR (Ian Sullivan )
- gain matching code updated, new unit tests added, small bug found and fixed
- responding to reviews on three tickets
- Also made simulated test dataset with variable PSFs and ran through his code; doesn't work as well, as expect. Extremely bad PSFs will drive model to divergence. Just did a simple ~grid of PSF sizes.
- what's it take to do HiTS?
- Meredith Rawls has tried running this, but makeCoaddTempExp is hanging–needs investigation.
- Ian Sullivan also has a smaller test DECam dataset
ap_
* development- Unknown User (cmorrison) PPDB seems to be running as expected, doing some quick checks to catch any null column handling issues.
- wrapping up the ticket, then on to flags.
- AOB/table round
- Eric thinking about matching to HiTS classified variables. Chris suggests could instantiate DIAObject table with those objects and let them collect