JP-3727: changed psfalign product from quadmodel to cubemodel #8747
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #8747 +/- ##
==========================================
+ Coverage 60.59% 60.63% +0.04%
==========================================
Files 372 372
Lines 38401 38372 -29
==========================================
- Hits 23269 23267 -2
+ Misses 15132 15105 -27 ☔ View full report in Codecov by Sentry. |
|
Regression tests are all accounted for: 18 of the 22 failures match the ones on the nightly runs here. The other four (two each for NIRCam and MIRI) are all directly testing the |
|
I am going to defer or hand off doing a review for this one, as I'm not an INS coronagraphy folks anymore. I would suggest that instead @aggle should review and concur for MIRI, and @juliengirard for NIRCam. |
|
Looks good to me as well! |
|
Okay, sounds like we have agreement. I'll go ahead and merge. Thanks all! |
Resolves JP-3727
Closes #8746
A previous pull request implemented a change that simplified the pipeline's coronagraphic data processing such that the alignment of the PSF to the science data is only done on the first science integration, then assumed to be the same for all subsequent integrations. Prior to this change, the shifted PSF was stored in a 4-dimensional data structure (QuadModel of shape n_sci_integrations, n_psf_integrations, n_rows, n_cols), which made sense when the shift was different on each science integration. Now, though, this model is identical over the zeroth axis and therefore redundant with itself. Furthermore, this model uses a lot of memory (see JP-3724), and is likely a (the?) cause of the out-of-memory failures reported on multiple datasets in JWSTDMS-921 (see comments on that ticket for additional datasets). This PR stores the shifted PSF as a CubeModel instead.
Checklist for PR authors (skip items if you don't have permissions or they are not applicable)
CHANGES.rstwithin the relevant release sectionHow to run regression tests on a PR