docs: JAX gradient audit README (status table + mesh verdicts)#88
Merged
Conversation
… verdicts Adds the per-likelihood gradient status table for the JAX autodiff audit (#87): FD-validated results for Sersic / linear Sersic / MGE / point-source source- plane / weak lensing, the re-confirmed Delaunay pure_callback-no-JVP hard error with the three-layer infeasibility write-up, and the rectangular-mesh staircase verdict (RectangularUniform exact; RectangularAdaptDensity at os_pix=1 is piecewise-constant in mass/shear — autodiff correct, smooth mass information destroyed by the empirical rank transform; os_pix>1 restores it). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XGDp54jKUd3kUziTF77pNu
… at os_pix=4 Both adaptive rectangular meshes FD-validated at production over-sampling: RectangularAdaptImage (reg.Adapt + AdaptImages + border relocator) <= ~1% on mass/shear, AdaptDensity <= ~3%; implications section updated — gradient-based mass inference works in the production config, only the os_pix=1 adaptive corner is unusable. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XGDp54jKUd3kUziTF77pNu
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Part of the JAX autodiff gradients audit — #87.
Scripts Changed
jax_profiling/gradient/README.md(new) — the audit's central artifact: per-likelihood gradient status table (all FD-validated results with dates + tolerances), the rectangular-mesh staircase verdict (RectangularUniformexact;RectangularAdaptDensityat os_pix=1 piecewise-constant in mass/shear — autodiff correct, smooth mass information destroyed by the empirical rank transform, os_pix>1 restores it; measured FD step-size table included), the three-layer Delaunay infeasibility write-up (pure_callbackno-JVP →custom_jvpzero-rule path), and the findings log.Companion PR (FD test scripts): PyAutoLabs/autolens_workspace_test#157 — see the audit issue for cross-references.
Validation
Documentation only; all numbers reproduced 2026-07-09 against current mains by the probes in
jax_profiling/gradient/and thejax_gradFD suite. Heart YELLOW acknowledged by the user at ship (pre-existing, unrelated).🤖 Generated with Claude Code
https://claude.ai/code/session_01XGDp54jKUd3kUziTF77pNu