Switch to use romanisim by yuedong0607 · Pull Request #100 · DukeCosmology/roman_imsim

yuedong0607 · 2026-04-28T21:04:40Z

This RP migrates the codebase to a new set of APIs in romanisim (here's an introduction of the new API), replacing all usages of galsim.roman. This change enables access to both CRDS and the roman-technical-information repository.

Current status:

Replaced all galsim.roman calls with romanisim APIs.
Updated the bandpass class to support spatially varying (per-SCA) throughputs, calculated from effective area files retrieved from roman-technical repo.
Added configurable options for individual instrumental effects in noise.py

TODO / Open Questions (for this PR):

Configuration schema: Finalize the format of configurable options in the YAML config. Some modules can take additional parameters for further configuration (see an example below for the nonlinearity module)

TODO / Open Questions (out of scope for this PR):

WCS: Currently still using the same implementation inherited from galsim.roman. Need to consider switching to use the romanisim.wcs APIs which is based on gwcs and leverages the CRDS distortion models.
PSF: Currently still using the same implementation inherited from galsim.roman. Need to consider switching to use the romanisim.psf APIs which supports both empirical PSF samples from CRDS, and STPSF models. However, both the empirical PSF and STPSF in romanisim.psf only support achromatic modeling at this moment.

from romanisim.models import Nonlinearity
nl = Nonlinearity(
        usecrds=True,
        reftype="inverselinearity",
        getdq=True,
        integralnonlinearity=None,
        metadata=None,
        image_mod=None,
        reffiles=None,
        saturation=None,
        Coeffs=None,
        gain=None,
)
nl.apply(img, electrons=False)

integralnonlinearity : object or None, optional
        Optional integral nonlinearity reference model. If provided, the class
        constructs channel-based lookup corrections that are added after the
        polynomial correction.

yuedong0607 · 2026-04-30T22:00:31Z

romanisim.models has not yet been included in the current release (v0.13.1) of romanisim. The merge needs to wait until the next release version.

yuedong0607 · 2026-05-13T20:10:56Z

I updated the romanisim dependency in pyproject.toml to point to the main branch, and all unit tests are now passing. I think the code is ready for review now. @arunkannawadi

arunkannawadi

I have reviewed everything expect for the changes in psf.py. Given the slowness you report, I think that needs some investigation.

yuedong0607 · 2026-06-02T07:51:26Z

I have reviewed everything expect for the changes in psf.py. Given the slowness you report, I think that needs some investigation.

It seems like the efficiency issue appears in both this branch and the integrated_coadd_sim branch (which doesn't use romanisim.models yet). I'm running some tests now.

yuedong0607 · 2026-06-16T06:55:45Z

I have reviewed everything expect for the changes in psf.py. Given the slowness you report, I think that needs some investigation.

I ran a round of code profiling last week and found that the performance issues previously observed in this branch were likely caused by some temporary problems on the DCC side. I've rebased the branch. It should now be ready for merging or any further review.

aguinot · 2026-06-23T02:50:29Z

+    def __init__(
+        self,
+        SCA=None,
+        WCS=None,
+        n_waves=None,
+        bpass=None,
+        extra_aberrations=None,
+        logger=None,
+    ):
+
+        logger = galsim.config.LoggerWrapper(logger)
+
+        if n_waves == -1:
+            if bpass.name == "W146":
+                n_waves = 10
+            else:
+                n_waves = 5
+
+        corners = [
+            galsim.PositionD(1, 1),
+            galsim.PositionD(1, models.parameters.n_pix),
+            galsim.PositionD(models.parameters.n_pix, 1),
+            galsim.PositionD(models.parameters.n_pix, models.parameters.n_pix),
+        ]
+        cc = galsim.PositionD(models.parameters.n_pix / 2, models.parameters.n_pix / 2)
+        tags = ["ll", "lu", "ul", "uu"]
+        self.PSF = {}
+        pupil_bin = 8
+        self.PSF[pupil_bin] = {}
+        for tag, SCA_pos in tuple(zip(tags, corners)):
+            self.PSF[pupil_bin][tag] = self._psf_call(
+                SCA, bpass, SCA_pos, WCS, pupil_bin, n_waves, logger, extra_aberrations
+            )
+        for pupil_bin in [4, 2, "achromatic"]:
+            self.PSF[pupil_bin] = self._psf_call(
+                SCA, bpass, cc, WCS, pupil_bin, n_waves, logger, extra_aberrations
+            )
+


Why is this added back here?
The PSFInterpolator is not supposed to be changed. It is only use as a constructor for specific interpolation scheme. What is being added here is already in the initPSF of the CornerPSFInterpolator see here.
Also this modification has nothing to do with this PR, is there is an issue/bug with the PSF implementation please make a separate PR to make it easier to track.

aguinot · 2026-06-23T02:52:34Z

+    def buildPSF(self, config, base, gsparams, logger):
+        """Build the PSF object.
+
+        For the Basic stamp type, this builds a PSF from the base['psf'] dict, if present,
+        else returns None.
+
+        Parameters:
+            config:     The configuration dict for the stamp field.
+            base:       The base configuration dict.
+            gsparams:   A dict of kwargs to use for a GSParams.  More may be added to this
+                        list by the galaxy object.
+            logger:     A logger object to log progress.
+
+        Returns:
+            the PSF
+        """
+        if base.get("psf", {}).get("type", "roman_psf") != "roman_psf":
+            return galsim.config.BuildGSObject(base, "psf", gsparams=gsparams, logger=logger)[0]
+
+        roman_psf = galsim.config.GetInputObj("roman_psf", config, base, "buildPSF")
+        psf = roman_psf.getPSF(self.pupil_bin, base["image_pos"])
+        return psf
+


Why is this beeing added back?
Same as the comment above, if there is a specific issue with the PSF it would be easier in a separate PR.

yuedong0607 force-pushed the switch_to_use_romanisim branch 2 times, most recently from 93c4cf2 to ac91f55 Compare April 30, 2026 21:46

yuedong0607 force-pushed the switch_to_use_romanisim branch from 88819ea to 414da08 Compare May 28, 2026 20:39

arunkannawadi reviewed May 29, 2026

View reviewed changes

Comment thread roman_imsim/bandpass.py Outdated

Comment thread roman_imsim/config/hack.yaml Outdated

Comment thread roman_imsim/noise.py Outdated

Comment thread roman_imsim/noise.py Outdated

yuedong0607 added 11 commits June 16, 2026 02:14

remove galsim.roman calls; calling from romanisim.models instead

80d7732

enabling the option for using per-SCA throughput

9e26d2d

add "per-effect" options to effects in noise class

f21df21

fix a typo

58de6b2

reformatted

e1d373a

formatted

0e00870

code formatted with black

9a8958b

black reformatted.

cf04fe4

specify bandnames=[name] in models.bandpass.getBandpasses

3b37405

remove print statement, change IPC to ipc_model in noise.py

9f1ed1c

fix a rebase error in psf.py, update default.yaml, remove hack.yaml

97ea3b9

yuedong0607 force-pushed the switch_to_use_romanisim branch from 8ebb3be to 97ea3b9 Compare June 16, 2026 06:24

aguinot reviewed Jun 23, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Switch to use romanisim#100

Switch to use romanisim#100
yuedong0607 wants to merge 11 commits into
mainfrom
switch_to_use_romanisim

yuedong0607 commented Apr 28, 2026 •

edited

Loading

Uh oh!

yuedong0607 commented Apr 30, 2026

Uh oh!

yuedong0607 commented May 13, 2026

Uh oh!

arunkannawadi left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

yuedong0607 commented Jun 2, 2026

Uh oh!

yuedong0607 commented Jun 16, 2026

Uh oh!

aguinot Jun 23, 2026

Uh oh!

aguinot Jun 23, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

yuedong0607 commented Apr 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

yuedong0607 commented Apr 30, 2026

Uh oh!

yuedong0607 commented May 13, 2026

Uh oh!

arunkannawadi left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

yuedong0607 commented Jun 2, 2026

Uh oh!

yuedong0607 commented Jun 16, 2026

Uh oh!

aguinot Jun 23, 2026

Choose a reason for hiding this comment

Uh oh!

aguinot Jun 23, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

yuedong0607 commented Apr 28, 2026 •

edited

Loading