DM-45856: Add new calibration pipelines for IsrTaskLSST to _ingredients #266
Conversation
|
See #267 for associated PR implementing the LATISS pipelines. |
| doBootstrap: false | ||
| doDeferredCharge: false | ||
| doBrighterFatter: false | ||
| doLinearize: true |
There was a problem hiding this comment.
General comment: Are these config options sorted in any fashion, and if not, can they be? Application order would be my preference (with doDiffNonLinearCorrection first), but even alphabetical would simplify things.
Also, can we use the same quoting (' vs ") throughout? I know the inconsistency is likely my fault, but it's throwing a lot of false positives in the diffs I'm doing to check the different pipeline versions.
There was a problem hiding this comment.
Argh, I meant to have everything double quoted and I will fix that.
As for ordering, I tried to keep it consistent but then didn't do a further pass. Which should I do ... application order or alphabetical?
| connections.outputData: "flatBootstrap" | ||
| calibrationType: "flat" | ||
| exposureScaling: InputList | ||
| scalingLevel: AMP |
There was a problem hiding this comment.
I assume this is AMP, where cpFlatLSST.yaml uses DETECTOR is due to this bootstrap version not having gains to normalize the amplifiers, so it's using the best guess from the normalization code? If so, a comment here might be useful so that it's clear that this difference is intentional.
There was a problem hiding this comment.
I don't understand the normalization conditions so this was at best accidental. This is used to find defects ... so I think AMP is appropriate? Unless you say it should be DETECTOR? Which should it be to find dark defects? None maybe.
There was a problem hiding this comment.
I think AMP is more correct here, but in thinking about the details of this, I don't think it matters. The defect code measures a clipped mean per amp to ensure each amp is on the same scale before doing the defect finding, so changing the flat scaling just changes the size of the offset subtracted by that clipped mean.
erykoff
force-pushed
the
tickets/DM-45856
branch
7 times, most recently
from
d321687
to
a923504
Compare
This removes the extraneous "exposure" because we do not want a new brighter fatter kernel for every exposure.
This avoids dimensionality clashes in existing repos; this will be fixed with DM-46439.
erykoff
force-pushed
the
tickets/DM-45856
branch
from
a923504
to
1a90571
Compare
This PR is the first implementation ticket for RFC-1044. The pipelines in this PR are not directly tested; the testing, in conjunction with the LATISS pipelines, is being done on DM-46356.