sbailey
commented
5 months ago Further details -- I think this would look like an override_NIGHT.yaml entry like
## 20230925 offset between telescope and dome screen so correct for fiberflat gradient
calibration:
autocalib_fiberflat:
solve_gradient: True which would propagate into desi_proc_joint_fit calling desispec.scripts.autocalib_fiberflat.main with an additional option --solve-gradient. Note that desispec.scripts.autocalib_fiberflat.main takes a list of strings as if it was constructing a command line for subprocess, and thus it needs a string '--solve-gradient'; it is not a function with key/value args that would need main(..., solve_gradient=True). We do not need to specify a reference night, nor does any reference night need to be processed before the night with the --solve-gradient override since the reference fiberflats are in $DESI_SPECTRO_CALIB itself.
Currently the override files are handled in PR #2187 desi_proc_night at the pipelining level. I think this is the first case where we need to be passing an option down into the algorithmic level.
akremin
Add hooks and override files to the pipeline to call
desi_autocalib_fiberflatwith the--solve-gradientoption for a list of nights with known gradients (likely due to calib screen misalignment with the telescope). From @rongpu and @sybenzvi in #2172 these nights are:20230925, 20231001, 20231002, 20231003, 20231004, 20231005, 20231006, 20231007, 20231008, 20231009, 20231010, 20231011, 20231012, 20231013, 20231014, 20231015, 20231203, 20231204, 20231205, 20231207, 20231214, 20231216, 20231217, 20231218, 20231219, 20231220, 20231221, 20231228, 20231229, 20231230, 20231231, 20240101, 20240102
Do this after PR #2187 is merged.