campaigns: hst_mge_cpu_pilot superseded by the phase-2a MGE/HST CPU cell

campaigns.yaml

@@ -162,15 +162,15 @@ tasks:
 - id: hst_mge_cpu_pilot
   campaign: setup-baseline
   title: Prepare and review the HST/MGE fp64 CPU pilot
-  status: needs-decision
+  status: superseded
   priority: high
   path: tasks/hst_mge_cpu_pilot.md
-  updated: '2026-10-13'
-  next: 'Review the Pulse collection choices (candidate input hashes and 31-worker resource plan verified without
-    execution) alongside project correctness audit #417. Resolve data origin/suitability, named host/stack/full
-    settings, witness/tolerances and background-load method. No freeze, compute or acceptance authorized. Post-/prm 2026-10-17: #418 merged (QUALIFICATION.md), #417 closed; record https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/hst-mge-pilot-qualification.md. Next bounded step: the synthetic witness harness that also fixes the background-load method and calibrates tolerances; still no collection, freeze or acceptance authorized.'
+  updated: '2026-10-25'
+  next: 'Superseded 2026-10-25 (human: adopt and close): the likelihood-lifecycle phase-2a imaging campaign
+    measured imaging/mge/hst jax_cpu_dense_fp64@hpc_ral_cpu at tag 2026.10.23.1 with parity match
+    (autolens_profiling#461). The pilot''s own protocol was not run. Do not reissue it.'
   issue: https://github.com/PyAutoLabs/autolens_profiling/issues/415
-  evidence: https://github.com/PyAutoLabs/autolens_profiling/pull/416
+  evidence: https://github.com/PyAutoLabs/autolens_profiling/pull/461
 - id: runtime_cell_single_jit_gpu_warmup
   campaign: point-source-source
   title: Runtime cells' A100 `single_jit` includes the post-compile warm-up — source-plane 0.642 ms vs a steady

dashboard.md

@@ -1,6 +1,6 @@
 # PyAutoPulse — profiling dashboard
 
-<!-- pulse-campaigns:534100f0568a525813aff5c2ac922f02789eb6251f61b94cb960a3b0b19be138 -->
+<!-- pulse-campaigns:3283e557de1d110da97abe4c9ffeaed914e3a35fa18b55feff3e32ebb760654d -->
 ## Profiling Check In
 
 <details><summary>Full check-in prompt</summary>
@@ -38,7 +38,6 @@ Reviewed 2026-10-17. Review the Pulse collection choices (candidate input hashes
 ### Active tasks
 
 - [Freeze the setup baseline specification, then collect and review evidence](https://github.com/PyAutoLabs/PyAutoPulse/blob/main/tasks/setup_baseline_campaign.md) — needs-decision: HST/MGE raw fixed-parameter CPU runner merged in #416; Pulse handoff #55 merged and Mind implementation closed. Resolve exact production host/stack/data, full settings, independent correctness witness/tolerances and background-load method. No collection, campaign freeze or acceptance authorized; global compile dispatch remains unavailable.
-- [Prepare and review the HST/MGE fp64 CPU pilot](https://github.com/PyAutoLabs/PyAutoPulse/blob/main/tasks/hst_mge_cpu_pilot.md) — needs-decision: Review the Pulse collection choices (candidate input hashes and 31-worker resource plan verified without execution) alongside project correctness audit #417. Resolve data origin/suitability, named host/stack/full settings, witness/tolerances and background-load method. No freeze, compute or acceptance authorized. Post-/prm 2026-10-17: #418 merged (QUALIFICATION.md), #417 closed; record https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/hst-mge-pilot-qualification.md. Next bounded step: the synthetic witness harness that also fixes the background-load method and calibrates tolerances; still no collection, freeze or acceptance authorized.
 
 </details>
 

receipts/fit.json

@@ -13,5 +13,5 @@
   "record_count": 31,
   "errors": [],
   "cached_from": null,
-  "refreshed_at": "2026-10-25T16:47:54Z"
+  "refreshed_at": "2026-10-25T16:49:29Z"
 }

receipts/lens-walltime.json

@@ -13,5 +13,5 @@
   "record_count": 56,
   "errors": [],
   "cached_from": null,
-  "refreshed_at": "2026-10-25T16:47:55Z"
+  "refreshed_at": "2026-10-25T16:49:30Z"
 }

receipts/lens.json

@@ -6,7 +6,7 @@
   "resolved_commit": "eef5edfe1c8b5b9bc8faa8f57beed7ae712d3130",
   "summary_path": "dashboard/catalogue.json",
   "source": "remote",
-  "fetched_at": "2026-10-25T16:47:53Z",
+  "fetched_at": "2026-10-25T16:46:41Z",
   "outcome": "ok",
   "schema": "profiling-summary",
   "version": 3,
@@ -18,5 +18,5 @@
     "superseded": 379,
     "archived": 99
   },
-  "refreshed_at": "2026-10-25T16:47:53Z"
+  "refreshed_at": "2026-10-25T16:49:27Z"
 }

snapshots/lens.json

@@ -4,7 +4,7 @@
   "github": "PyAutoLabs/autolens_profiling",
   "summary_path": "dashboard/catalogue.json",
   "commit": "eef5edfe1c8b5b9bc8faa8f57beed7ae712d3130",
-  "fetched_at": "2026-10-25T16:47:53Z",
+  "fetched_at": "2026-10-25T16:46:41Z",
   "summary": {
     "archive_record_count": 99,
     "comparison_policy": {

state.json

@@ -4,7 +4,7 @@
   "repo": "PyAutoPulse",
   "status": "yellow",
   "headline": "3 projects · 766 records · 0 cached",
-  "updated": "2026-10-25T16:47:55Z",
+  "updated": "2026-10-25T16:46:46Z",
   "pages_url": "https://pyautolabs.github.io/PyAutoPulse/",
   "items": [
     {

tasks/hst_mge_cpu_pilot.md

@@ -3,7 +3,7 @@
 Campaign: setup-baseline
 Task: hst_mge_cpu_pilot
 Reviewed: 2026-10-13
-Status: needs-decision
+Status: superseded
 
 The human approved preparation in this chat ("ok go", 2026-10-13). This document
 proposes the experiment; it does not freeze a specification or authorize compute,
@@ -218,3 +218,27 @@ was authorized. This focused preparation does not advance the all-campaign
 - The design (`scripts/misc/jax_compile/QUALIFICATION.md`) derives what the pilot computes, proposes an independent operator + scipy NNLS witness, lists the identities to capture and gives candidate tolerances marked unaccepted. No contradiction with the #416 runner.
 - Carried, not fixed: no background-load method proposed; the 1e-8 relative image L2 budget is likely unattainable against the packaged solver's inactive-column flux allowance, so synthetic-harness calibration must be allowed to revise tolerances; correctness-process feasibility is deferred, so the 32-process / 16 CPU-hour budget stays unfinalized. Witness placement in autolens_workspace_test with a project adapter needs the human's confirmation.
 - No collection, campaign freeze or acceptance authorized.
+
+## Superseded by phase 2a — 2026-10-25
+
+The human chose to adopt the likelihood-lifecycle phase-2a imaging campaign's MGE/HST
+CPU cell as this pilot's answer and close the pilot ("that's what I meant, so do it",
+2026-10-25). The pilot's own protocol was **not** run.
+
+- Answering cell: `results/imaging/mge/hst/jax_cpu_dense_fp64/hpc_ral_cpu/runtime.json`
+  in autolens_profiling, with `breakdown.json`, `fit.png` and `parity.json` (`match`
+  against the eager NumPy reference) beside it. It was measured on RAL `ral` CPU from a
+  private checkout at tag 2026.10.23.1 (batch 1 array 412925) and merged in
+  [autolens_profiling#461](https://github.com/PyAutoLabs/autolens_profiling/pull/461).
+  Campaign issue: [autolens_profiling#456](https://github.com/PyAutoLabs/autolens_profiling/issues/456).
+- How it differs from this proposal: the phase-2a cell runs the standard
+  production MGE at the instrument's seeded prior draw, not the 20+20 Gaussian fixed
+  instance. It records single-jit medians and vmap per-call times, with compile time
+  reported separately, not the cold / warm / persistent-cache stages. It applies no
+  background-load qualification: the dashboard refuses rows whose load average exceeded 8,
+  and the catalogue and coverage still count them. Correctness comes from the
+  per-cell parity witness, not the operator + scipy NNLS witness in QUALIFICATION.md.
+- The open questions above (persistent-cache auditing, background-load method, tolerance
+  calibration) are not answered by this adoption. If they matter again, open a new task;
+  do not reissue this pilot.
+