From 657f9d05a07c6f927f9b8a98bdaaf3f8468611d3 Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 18 Aug 2026 13:50:41 +0000 Subject: [PATCH] record: inference-programme-ledger shipped (autolens_profiling#136+#137) Ledger merged to autolens_profiling main; duplicate PR#135 reconciled via #137 and closed; issue #134 closed. Rich record notes the duplicate-work trap (task started from two devices) and the stale laptop worktree/branch for the next repo_cleanup pass. Co-Authored-By: Claude Claude-Session: https://claude.ai/code/session_018dV9h5h76hZ6trGZKWJFHe --- active.md | 13 ---- active/inference_programme_ledger.md | 36 ---------- .../2026/08/inference-programme-ledger.md | 69 +++++++++++++++++++ complete/index.md | 3 +- 4 files changed, 71 insertions(+), 50 deletions(-) delete mode 100644 active/inference_programme_ledger.md create mode 100644 complete/2026/08/inference-programme-ledger.md diff --git a/active.md b/active.md index 213a55ea..538d0ce8 100644 --- a/active.md +++ b/active.md @@ -1,18 +1,5 @@ # Active Tasks -## inference-programme-ledger -- issue: https://github.com/PyAutoLabs/autolens_profiling/issues/134 -- status: pr-open (https://github.com/PyAutoLabs/autolens_profiling/pull/135) -- worktree: ~/Code/PyAutoLabs-wt/inference-programme-ledger -- repos: - - autolens_profiling: feature/inference-programme-ledger -- prompt: active/inference_programme_ledger.md -- note: docs-only (results/notes/inference/ PROGRAMME.md + DECISIONS.md + LITERATURE.md). - The mge-lane-death claim it was recorded as disjoint from is gone — archived - 2026-08-18 as superseded (`complete/2026/08/mge-lane-death.md`); its - `research/mge-lane-death` branch was never created. - This is the canonical copy of the 2026-08-17 human-approved inference programme. - ## stored-sample-reconstruction-guard - issue: https://github.com/PyAutoLabs/PyAutoFit/issues/1486 - status: library-dev — WORKSPACE HALF SHIPPED; the PyAutoFit hardening (#1486) is what remains diff --git a/active/inference_programme_ledger.md b/active/inference_programme_ledger.md deleted file mode 100644 index d307047d..00000000 --- a/active/inference_programme_ledger.md +++ /dev/null @@ -1,36 +0,0 @@ -# Commit the inference-methods programme plan + knowledge ledger into autolens_profiling - -Type: docs -Target: autolens_profiling -Repos: -- @autolens_profiling -Difficulty: small -Autonomy: supervised -Priority: high -Status: draft - -The 2026-08-17 inference-methods planning pass (human-approved same day, with -notes) produced a full phased R&D programme: state of play, phases 0-13 with -decision gates A-F, benchmark/result schema v2, method-card knowledge structure, -source-change list, risk register, critical path, and a hypothetical decision -tree. Human review notes: (1) PositionsLH bug — verified, fix shipped as -PyAutoLens#699/PR#700; (2) BlackJAX NS "was fastest or comparable on MGE and may -scale better — worth rerunning" (Phase 2 stands); (3-6) approved as planned. - -Per the programme's own knowledge rules, the canonical copy must live in -autolens_profiling, not in chat history or an artifact. - -Task: create `results/notes/inference/` containing: - -- `PROGRAMME.md` — the full plan (markdown adaptation of the approved artifact), - maintained as phases execute. -- `DECISIONS.md` — append-only gate log, seeded with the 2026-08-17 plan - approval + review notes and the CP-1 completion (PositionsLH fix PR). -- `LITERATURE.md` — the external references gathered during planning (blackjax - 1.6 nested sampling merge, NSS paper, GIGA-Lens/Herculens, NUTS diagnostics, - informed-start SMC, MCLMC/MAMS, Pathfinder, proximal smoothing), one lesson - line each. - -No profiling runs, no scripts, no source changes — documentation only. Keep -`ruff check .` / `ruff format --check .` green (no .py files added) and do not -touch the auto-generated README tables. diff --git a/complete/2026/08/inference-programme-ledger.md b/complete/2026/08/inference-programme-ledger.md new file mode 100644 index 00000000..ac3a324c --- /dev/null +++ b/complete/2026/08/inference-programme-ledger.md @@ -0,0 +1,69 @@ +Committed the human-approved 2026-08-17 inference-methods programme into its +canonical home, `autolens_profiling/results/notes/inference/`, per the +programme's own knowledge rules (no knowledge lives only in chat histories or +artifacts). Documentation only — no runs, no scripts, no source changes. + +- **PRs:** autolens_profiling#136 (merged `3be8158` — PROGRAMME.md + + DECISIONS.md + LITERATURE.md, adapted from the approved artifact, with a + maintained phase/gate state table and CP-1 marked complete) and + autolens_profiling#137 (merged `5ce0243` — folded in the laptop session's + richer content; closed issue #134). +- **Duplicate-work trap:** the task was started twice. A laptop session had + already opened autolens_profiling#135 (`feature/inference-programme-ledger`, + registered in active.md) when a mobile/cloud session redid the task on + `claude/inference-methods-autolens-w86nz9` and merged #136 without checking + the registry first. The versions were complementary — #136 had the fuller + PROGRAMME body (995 vs 558 lines), #135 the richer CP-1 evidence + (test-pin values, 538-test suite, fence-slope-halving science impact, + truth-bars-unaffected note), the deeper LITERATURE set (~25 refs incl. the + optimizer-termination section), and the Future follow-ups section. #137 + reconciled them; #135 closed unmerged, nothing lost. Lesson: check + active.md's PR/status line before re-issuing an in-flight task from another + device. +- **Leftovers:** laptop worktree `~/Code/PyAutoLabs-wt/inference-programme-ledger` + and remote branch `feature/inference-programme-ledger` are stale — next + repo_cleanup pass. +- **What the ledger seeds:** DECISIONS.md carries the plan approval (verbatim + review notes) + CP-1 completion (PositionsLH accumulation fix, + PyAutoLens#699/PR#700). Next programme steps per PROGRAMME.md's critical + path: CP-2 blackjax ≥1.6.2 upgrade + mainline NSS smoke (laptop GPU), then + CP-3 Prodigy reliability scan; Phase 0 (b)(c)(e) outstanding. + +## Original prompt + +# Commit the inference-methods programme plan + knowledge ledger into autolens_profiling + +Type: docs +Target: autolens_profiling +Repos: +- @autolens_profiling +Difficulty: small +Autonomy: supervised +Priority: high +Status: draft + +The 2026-08-17 inference-methods planning pass (human-approved same day, with +notes) produced a full phased R&D programme: state of play, phases 0-13 with +decision gates A-F, benchmark/result schema v2, method-card knowledge structure, +source-change list, risk register, critical path, and a hypothetical decision +tree. Human review notes: (1) PositionsLH bug — verified, fix shipped as +PyAutoLens#699/PR#700; (2) BlackJAX NS "was fastest or comparable on MGE and may +scale better — worth rerunning" (Phase 2 stands); (3-6) approved as planned. + +Per the programme's own knowledge rules, the canonical copy must live in +autolens_profiling, not in chat history or an artifact. + +Task: create `results/notes/inference/` containing: + +- `PROGRAMME.md` — the full plan (markdown adaptation of the approved artifact), + maintained as phases execute. +- `DECISIONS.md` — append-only gate log, seeded with the 2026-08-17 plan + approval + review notes and the CP-1 completion (PositionsLH fix PR). +- `LITERATURE.md` — the external references gathered during planning (blackjax + 1.6 nested sampling merge, NSS paper, GIGA-Lens/Herculens, NUTS diagnostics, + informed-start SMC, MCLMC/MAMS, Pathfinder, proximal smoothing), one lesson + line each. + +No profiling runs, no scripts, no source changes — documentation only. Keep +`ruff check .` / `ruff format --check .` green (no .py files added) and do not +touch the auto-generated README tables. diff --git a/complete/index.md b/complete/index.md index 3c51183d..6128c913 100644 --- a/complete/index.md +++ b/complete/index.md @@ -6,7 +6,7 @@ Token-light navigation over the finished-work records (schema: only then grep a dated bucket. Curators: edit the band between the CURATED markers; everything below GENERATED is rebuilt. -1005 records across 7 buckets. +1006 records across 7 buckets. ## Highlights @@ -67,6 +67,7 @@ _(curate hard-won records here — survives regeneration.)_ - [hygiene-detail-flag](2026/08/hygiene-detail-flag.md) - [hygiene-gone-ref-counter](2026/08/hygiene-gone-ref-counter.md) — the prompt was never issued — the defects were fixed under other issues, see below - [hygiene-howto-refs-docstrings](2026/08/hygiene-howto-refs-docstrings.md) +- [inference-programme-ledger](2026/08/inference-programme-ledger.md) - [interferometer-delaunay-flaky-fitexception](2026/08/interferometer-delaunay-flaky-fitexception.md) - [interferometer-start-here-integrate-oom](2026/08/interferometer-start-here-integrate-oom.md) - [intra-family-dep-floors](2026/08/intra-family-dep-floors.md)