Repository navigation
feat: Preloads API for reusing channel-invariant inversion quantities - #344
Merged
Merged
Conversation
Add an opt-in `Preloads` mechanism: precompute an invariant inversion quantity once and inject it so repeated evaluations reuse it instead of rebuilding it. - New `aa.AbstractPreloads` / `aa.PreloadsInterferometer` containers (autoarray/preloads/) with an optional `curvature_matrix` field. - The inversion stack (`AbstractInversion`, the interferometer inversions, and the `inversion_from` / `inversion_interferometer_from` factory) gains an optional `preloads=None` argument, threaded as a single object. - The sparse interferometer inversion returns `preloads.curvature_matrix` (`F = LᵀW̃L`) directly when provided, skipping the dominant rebuild. Default behaviour is unchanged (`preloads=None`). The primary use is a shared / combined likelihood (e.g. a datacube FactorGraphModel) where invariance is explicit; an earlier preload system was removed for being bug-prone, so the docstrings make the opt-in, caller-owns-invariance contract explicit. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Extend AbstractPreloads / PreloadsInterferometer with an optional mapper_galaxy_dict field, so a pre-built mapper (carrying the expensive Delaunay triangulation) can be preloaded and reused alongside the curvature matrix. Stored opaquely; the consumer populates/interprets it. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Collaborator
Author
|
Updated ( |
This was referenced Jun 8, 2026
Merged
4 tasks done
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds an opt-in
Preloadsmechanism to the inversion stack: a channel-invariant inversion quantity (starting with the curvature matrixF = LᵀW̃L, the dominant inversion-setup cost) can be computed once and injected so repeated evaluations reuse it instead of rebuilding it. This is the PyAutoArray half of the datacube cross-Analysisshared-state feature (PyAutoLens#565 / theanalysis_shared_stateepic): when every spectral channel of a datacube shares one lens model,Fis identical for all of them and is now built once and reused.The primary, safe home for preloads is a shared / combined likelihood, where invariance is explicit and easy to verify. An earlier preload system was removed because it was bug-prone and hard to know when a quantity was safe to preload; the
AbstractPreloadsdocstring records that history and makes the opt-in, caller-owns-invariance contract explicit, and preloads are deliberately not applied as standard.API Changes
All additive and backward compatible — no removals or renames.
aa.AbstractPreloadsandaa.PreloadsInterferometer(curvature_matrix=None)— containers for preloaded inversion quantities (newautoarray/preloads/package).AbstractInversion, the interferometer inversions, and theinversion_from/inversion_interferometer_fromfactory gain an optionalpreloads=Noneargument.InversionInterferometerSparse.curvature_matrixnow returnspreloads.curvature_matrixdirectly when a preload is supplied (otherwise computes it as before).See full details below.
Test Plan
python -m pytest test_autoarray/— full suite passes (841 tests)test_autoarray/inversion/inversion/interferometer/test_interferometer.py::test__preloads_interferometer__curvature_matrix_returned_directly_and_skips_rebuild: a sentinel preload is returned verbatim (theFbuild is skipped), an emptyPreloadsInterferometer()falls back to the standard build, and preloading the realFleaves the per-channeldata_vectorunchanged.Full API Changes (for automation & release notes)
Added
autoarray.preloads.abstract.AbstractPreloads(curvature_matrix=None)(exported asaa.AbstractPreloads).autoarray.preloads.interferometer.PreloadsInterferometer(curvature_matrix=None)(exported asaa.PreloadsInterferometer).Changed Signature (backward compatible — new optional kwarg, default
None)AbstractInversion.__init__(... , preloads=None)AbstractInversionInterferometer.__init__(... , preloads=None)InversionInterferometerSparse.__init__(... , preloads=None)inversion_from(... , preloads=None)inversion_interferometer_from(... , preloads=None)Changed Behaviour
InversionInterferometerSparse.curvature_matrixreturnspreloads.curvature_matrixwhen a preload with a non-Nonecurvature_matrixis supplied; otherwise unchanged.Migration
preloads=aa.PreloadsInterferometer(curvature_matrix=F)to the inversion / factory.🤖 Generated with Claude Code