Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: obs_subaru
-
Labels:None
-
Story Points:3
-
Epic Link:
-
Sprint:DRP F18-4
-
Team:Data Release Production
Description
Add reference catalog defaults to at least obs_subaru to facilitate running jointcal on the HSC biweeklies.
John Parejko I assigned it to you, but I can just as easily do it if one of you sends me the command to run jointcal on the RC2 dataset so I can test that the configs make it do what we expect.
Attachments
Activity
The config override I used in DM-15521 was
config.astrometryRefObjLoader.ref_dataset_name='ps1_pv3_3pi_20170110'
|
config.photometryRefObjLoader.ref_dataset_name='ps1_pv3_3pi_20170110'
|
config.astrometryRefObjLoader.filterMap={'B': 'g', 'r2': 'r', 'N1010': 'z', 'N816': 'i', 'I': 'i', 'N387': 'g', 'i2': 'i', 'R': 'r', 'N921': 'z', 'N515': 'g', 'V': 'r'}
|
config.photometryRefObjLoader.filterMap={'B': 'g', 'r2': 'r', 'N1010': 'z', 'N816': 'i', 'I': 'i', 'N387': 'g', 'i2': 'i', 'R': 'r', 'N921': 'z', 'N515': 'g', 'V': 'r'}
|
but as said, probably better to be exactly the same as in the single frame processing such as loading filterMap.py like here in obs_subaru.
We do have /datasets/refcats/htm/gaia_DR1_v1/ on LSST machines. Do you prefer that?
One command for the test run can be:
jointcal.py /datasets/hsc/repo --calib /datasets/hsc/repo/CALIB --rerun RC/w_2018_34/DM-15517:private/user/jointcal --id ccd=0..8^10..103 visit=26024^26028^26032^26036^26044^26046^26048^26050^26058^26060^26062^26070^26072^26074^26080^26084^26094 filter=HSC-G tract=9615
|
Hsin-Fang Chiang
added a comment - The config override I used in DM-15521 was
config.astrometryRefObjLoader.ref_dataset_name='ps1_pv3_3pi_20170110'
config.photometryRefObjLoader.ref_dataset_name='ps1_pv3_3pi_20170110'
config.astrometryRefObjLoader.filterMap={'B': 'g', 'r2': 'r', 'N1010': 'z', 'N816': 'i', 'I': 'i', 'N387': 'g', 'i2': 'i', 'R': 'r', 'N921': 'z', 'N515': 'g', 'V': 'r'}
config.photometryRefObjLoader.filterMap={'B': 'g', 'r2': 'r', 'N1010': 'z', 'N816': 'i', 'I': 'i', 'N387': 'g', 'i2': 'i', 'R': 'r', 'N921': 'z', 'N515': 'g', 'V': 'r'}
but as said, probably better to be exactly the same as in the single frame processing such as loading filterMap.py like here in obs_subaru.
We do have /datasets/refcats/htm/gaia_DR1_v1/ on LSST machines. Do you prefer that?
One command for the test run can be:
jointcal.py /datasets/hsc/repo --calib /datasets/hsc/repo/CALIB --rerun RC/w_2018_34/DM-15517:private/user/jointcal --id ccd=0..8^10..103 visit=26024^26028^26032^26036^26044^26046^26048^26050^26058^26060^26062^26070^26072^26074^26080^26084^26094 filter=HSC-G tract=9615
We do have /datasets/refcats/htm/gaia_DR1_v1/ on LSST machines. Do you prefer that?
Lets stick with PS1 for now, and see what gaia does as a different test.
I found ref cats and filterMaps overrides in two other obs_* packages, that we'd conceivably want to run jointcal on: obs_lsstCam, obs_decam.
Would you review?
Dominique Boutigny, at a quick glance I couldn't tell which reference catalog is your default in https://github.com/lsst/obs_cfht/blob/master/config/processCcd.py
Hmm. Needs a little more.
The unit tests do:
self.config.astrometryRefObjLoader.retarget(LoadAstrometryNetObjectsTask)
but the overrides in the obs* packages get used in the unit tests (via parseAndRun).
(I understand Dominique's note in obs_cfht now)
=================================== FAILURES ===================================
|
JointcalTestDECAM.test_jointcalTask_2_visits_constrainedAstrometry_no_photometry
|
[gw3] linux -- Python 3.6.6 /j/ws/stack-os-matrix/centos-7.devtoolset-6.py3/lsstsw/miniconda/envs/lsst-scipipe/bin/python3.6
|
|
|
self = <test_jointcal_decam.JointcalTestDECAM testMethod=test_jointcalTask_2_visits_constrainedAstrometry_no_photometry>
|
|
|
def test_jointcalTask_2_visits_constrainedAstrometry_no_photometry(self):
|
relative_error, pa1, metrics = self.setup_jointcalTask_2_visits_constrainedAstrometry_no_photometry()
|
|
> self._testJointcalTask(2, relative_error, self.dist_rms_absolute, pa1, metrics=metrics)
|
|
|
tests/test_jointcal_decam.py:109:
|
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
|
tests/jointcalTestBase.py:134: in _testJointcalTask
|
result = self._runJointcalTask(nCatalogs, caller, metrics=metrics)
|
tests/jointcalTestBase.py:180: in _runJointcalTask
|
result = jointcal.JointcalTask.parseAndRun(args=args, doReturnResults=True, config=self.config)
|
../../stack/Linux64/pipe_base/16.0-11-g9fe0e56+9/python/lsst/pipe/base/cmdLineTask.py:598: in parseAndRun
|
parsedCmd = argumentParser.parse_args(config=config, args=args, log=log, override=cls.applyOverrides)
|
../../stack/Linux64/pipe_base/16.0-11-g9fe0e56+9/python/lsst/pipe/base/argumentParser.py:634: in parse_args
|
self._applyInitialOverrides(namespace)
|
../../stack/Linux64/pipe_base/16.0-11-g9fe0e56+9/python/lsst/pipe/base/argumentParser.py:832: in _applyInitialOverrides
|
namespace.config.load(filePath)
|
../../stack/Linux64/pex_config/16.0-3-g9645794+3/python/lsst/pex/config/config.py:543: in load
|
self.loadFromStream(stream=code, root=root)
|
../../stack/Linux64/pex_config/16.0-3-g9645794+3/python/lsst/pex/config/config.py:563: in loadFromStream
|
exec(stream, {}, local)
|
../../stack/Linux64/obs_decam/tickets.DM-15606-g8e0a9334c4/config/jointcal.py:1: in <module>
|
config.astrometryRefObjLoader.ref_dataset_name = 'ps1_pv3_3pi_20170110'
|
../../stack/Linux64/pex_config/16.0-3-g9645794+3/python/lsst/pex/config/configurableField.py:122: in __setattr__
|
self._value.__setattr__(name, value, at=at, label=label)
|
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
|
|
|
self = lsst.meas.algorithms.loadReferenceObjects.LoadReferenceObjectsConfig(pixelMargin=300, defaultFilter='', filterMap={})
|
attr = 'ref_dataset_name', value = 'ps1_pv3_3pi_20170110'
|
at = [StackFrame(<string>, 1, <module>), StackFrame(<string>, 1539, <module>), StackFrame(<string>, 1534, serve), StackFram..., 1047, serve), StackFrame(<string>, 259, integrate_as_primary_thread), StackFrame(<string>, 277, _perform_spawn), ...]
|
label = 'assignment'
|
|
|
def __setattr__(self, attr, value, at=None, label="assignment"):
|
"""!Regulate which attributes can be set
|
|
Unlike normal python objects, Config objects are locked such
|
that no additional attributes nor properties may be added to them
|
dynamically.
|
|
Although this is not the standard Python behavior, it helps to
|
protect users from accidentally mispelling a field name, or
|
trying to set a non-existent field.
|
"""
|
if attr in self._fields:
|
if at is None:
|
at = getCallStack()
|
# This allows Field descriptors to work.
|
self._fields[attr].__set__(self, value, at=at, label=label)
|
elif hasattr(getattr(self.__class__, attr, None), '__set__'):
|
# This allows properties and other non-Field descriptors to work.
|
return object.__setattr__(self, attr, value)
|
elif attr in self.__dict__ or attr in ("_name", "_history", "_storage", "_frozen", "_imports"):
|
# This allows specific private attributes to work.
|
self.__dict__[attr] = value
|
else:
|
# We throw everything else.
|
> raise AttributeError("%s has no attribute %s" % (_typeStr(self), attr))
|
E AttributeError: lsst.meas.algorithms.loadReferenceObjects.LoadReferenceObjectsConfig has no attribute ref_dataset_name
|
Chris Morrison [X] is in the process of building indexed refcats for the validation datasets, so I should be be able to switch testdata_jointcal to use those once that's done. That might fix the above problem? Or do you have another suggestion about how to have the unittests not get overridden? They're supplying a Config object, so I'm somewhat surprised that it's getting overridden (though I don't know what the hierarchy is there).
Per DM-15713, it looks like we want to go with a 7th order visit polynomial for HSC astrometry. This means adding the following to the hsc jointcal config (modify the comment as you see fit):
# Testing in DM-15713 found that a 7th order visit polynomial markedly improved the astrometric residuals on HSC.
|
self.config.astrometryVisitOrder = 7
|
The NAOJ team reported once that a 9th order polynomial was necessary to model the distortion in HSC, which turns on hard at the edge of the FoV.
Paul Price
added a comment - The NAOJ team reported once that a 9th order polynomial was necessary to model the distortion in HSC, which turns on hard at the edge of the FoV. The NAOJ team reported once that a 9th order polynomial was necessary to model the distortion in HSC, which turns on hard at the edge of the FoV.
Given that jointcal is exceeding meas_mosaic on the SRD metrics with both a 5th and 7th order, I'd rather not go to higher order without some further measurements to show its necessary. That can come later.
I'm going to take this over for a bit to see if I can fix the jointcal failures it introduces.
Yusra AlSayyad: I came up with a possible solution, implemented for the jointcal decam tests. Can you please take a look and tell me if you'd be happy with this? If so, I'll do it for the other tests too.
https://github.com/lsst/jointcal/compare/tickets/DM-15606
It's not great (it turns lines of actual code into long strings), but it means I don't have to re-implement parts of parseAndRun to build up the dataId list, etc.
A configfile solution is on u/yusra/DM-15606 in jointcal and passes Jenkins (https://ci.lsst.codes/job/stack-os-matrix/28702). I'll open a PR if you'd be up for reviewing. Before merging, I would put it on tickets/DM-15606 and squash the commits.
I added the config.astrometryVisitOrder = 7 to the obs_subaru PR too.
Quick skim suggests this is much better than my attempt. I'd be happy to review, and will expedite it: rename the ticket branch now (feel free to wipe out my branch above) and do whatever squashing you want, and I'll review the PR asap.
Commits squashed, PR opened, and https://ci.lsst.codes/job/stack-os-matrix/28714 succeeded.
One tiny comment, otherwise this looks very good. Thanks for coming up with it!
Thank you John Parejko
Hsin-Fang Chiang has the scripts for running jointcal on RC2.
And I don't know what appropriate defaults are: I think the correct thing is to go with whatever Hsin-Fang Chiang did for her recent run. There's no reason jointcal can't use the same reference catalog defaults as single frame processing for both the astrometry and photometry refcats, for now. If we have gaia, we should maybe use that for astrometry (it doesn't have appropriate photometry yet).