Shelve superseded group4 MGE search benchmark - #205
Merged
Conversation
The group4 MGE search benchmark (parked.md :: group4-mge-search-benchmark, autolens_profiling#82) is superseded by the 2026-08-17 inference-methods programme (active/inference_programme_ledger.md, autolens_profiling#134), which now owns sampler-benchmark direction. Its first half already shipped (code + first GPU results merged as autolens_profiling#83); the remaining gradient-family sweep is not carried forward as scoped. Like the sampler wave, it showed on the dashboard twice — once under In flight (the prompt still sat in active/) and once under Parked. - move active/research_profiling_experiment_in_the_autolens_pr.md -> complete/archive/shelved/, with a SUPERSEDED header preserving the full parked context (remaining sweep + GPU env, post-#84 script paths, sweep driver invocation, compile-cost knobs) that the deleted parked.md entry carried - drop the parked.md entry - regenerate dashboard.md (in flight 9 -> 8, parked 3 -> 2) Dated completion records that name the task (complete/2026/08/active-prompt-orphan-triage.md, complete/2026/07/profiling-mirror-taxonomy.md) are left as history; the restructure the latter waited on has since landed (#84). lifecycle.py check: OK Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016X8zrRQuau8VoGwx3i45tf
Jammy2211
added a commit
that referenced
this pull request
Aug 18, 2026
The MGE multi-start lane investigation (active.md :: mge-lane-death, autolens_profiling#128) is done in substance: the cause was found on 2026-08-15 and every follow-up it owed has since shipped. The remaining confirmation legs (GPU / float64 / multi-seed re-measurement of the 62%) are superseded and not carried forward as scoped. Unlike the two shelvings that preceded it (#204, #205), this one earns a dated completion record rather than complete/archive/shelved/ — the work produced a cause, a refuted hypothesis, a counter-finding and six merged PRs, and five existing records already cite the investigation. - lifecycle.py record mge-lane-death --date 2026-08-18: writes complete/2026/08/mge-lane-death.md (cause, evidence, the masked-ell_comps counter-finding, the survival-integral trap, the shipped downstream PRs, and what is not carried forward), folds in active/mge_lane_death.md, and drops the active.md section - repoint the inference-programme-ledger disjointness note at the record; its research/mge-lane-death branch was never created - regenerate complete/index.md and dashboard.md (in flight 8 -> 7) Tracking issue autolens_profiling#128 is still open and wants closing by hand — that repo is out of this session's scope. lifecycle.py check: OK lifecycle.py index --check: OK Claude-Session: https://claude.ai/code/session_013xP1c7XRVQJzxoWAoH92GQ Co-authored-by: Claude <noreply@anthropic.com>
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
2 participants
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.
Follow-up to #204, same treatment for the next superseded dashboard row.
The group4 MGE search benchmark (
parked.md :: group4-mge-search-benchmark, autolens_profiling#82) is superseded by the 2026-08-17 human-approved inference-methods programme (active/inference_programme_ledger.md, autolens_profiling#134), which now owns sampler-benchmark direction. Its first half already shipped — code + first GPU results merged as autolens_profiling#83 — and the remaining gradient-family sweep is not carried forward as scoped.Like the sampler wave, it was showing on
dashboard.mdtwice: once under In flight (the prompt still sat inactive/) and once under Parked.Changes
active/research_profiling_experiment_in_the_autolens_pr.md→complete/archive/shelved/, the repo's home for deferred/shelved prompts, whichlifecycle.pycheck/indexdeliberately skip.SUPERSEDEDheader preserving everything the deletedparked.mdentry carried — the remaining sweep (prodigy / lion / adabelief / prodigy_autoconv + Nautilus anchor) and its GPU env, the post-Organism docs sweep: promote Nerves to 7th organ + PyAutoBrain AGENTS.md diet + stale-ref fixes #84 script paths (scripts/cluster/searches/<sampler>/mge.py), thesweep.py --only <sampler>/group/mgeinvocation and its internal group-cell keying, and the compile-cost knobs (_GROUP4_MGE_TOTAL_GAUSSIANS,_MULTI_START_N_STARTS). That detail existed nowhere else.parked.mdsection.dashboard.md: in flight 9 → 8, parked 3 → 2.Left alone deliberately
complete/2026/08/active-prompt-orphan-triage.mdandcomplete/2026/07/profiling-mirror-taxonomy.mdname the task — both are dated completion records describing what was true at the time. Records are history; editing them is drift. The restructure the latter deferred behind this claim has since landed (autolens_profiling#84), so nothing live is blocked.Verification
python3 scripts/lifecycle.py check→lifecycle check: OKGenerated by Claude Code