From 9ea5bddb72cad0e15d93ca05d39146f60c7fc0c4 Mon Sep 17 00:00:00 2001 From: Jammy2211 Date: Mon, 19 Oct 2026 18:19:20 +0100 Subject: [PATCH 1/2] =?UTF-8?q?insight:=20hst-imaging-mass-samplers=202026?= =?UTF-8?q?-10-19=20clean-up=20=E2=80=94=20412541/412565=20cancelled=20(hu?= =?UTF-8?q?man),=20flat=20layout=20applied,=20wave-1=20attempt=201=20set?= =?UTF-8?q?=20aside?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Focused record (not a full sweep; last_checkin unchanged). 412539 COMPLETED; 412541 and 412565 scancelled ~15:51Z on the human's decision with partial stages left on disk; autolens_inference#36 merged (c3ca4de) and flatten_cleanslate.py --apply run on RAL and the laptop mirror (1974 files identical); wave-1 attempt-1 experiments/experiment_sources/failed rows moved to the aside folders. Keeping prepared/ recorded as a proposal pending human confirmation. No compute submitted. Co-Authored-By: Claude Opus 5.5 --- campaigns.yaml | 34 +++++----- checkins/2026-10-19T171637Z.md | 99 ++++++++++++++++++++++++++++++ dashboard.html | 2 +- dashboard.md | 10 +-- tasks/hst_mass_base_light_modes.md | 12 ++++ tasks/hst_mass_samplers_wave1.md | 3 +- 6 files changed, 136 insertions(+), 24 deletions(-) create mode 100644 checkins/2026-10-19T171637Z.md diff --git a/campaigns.yaml b/campaigns.yaml index a5a1d61..f89a080 100644 --- a/campaigns.yaml +++ b/campaigns.yaml @@ -37,15 +37,15 @@ campaigns: - id: hst-imaging-mass-samplers title: 'HST imaging, clean slate: Nautilus base on A100, jax_cpu and numba_cpu at a pinned release, mass stage with linear lens light and with it fixed, then other samplers on the mass stage' - status: running - updated: '2026-10-18' - next: '(1) Human decision on filing a Mind bug task for the jax_gpu sparse fixed-light mass_total[1] divergence (both seeds) and reproducing it on the laptop from the exported prepared problem, before any CPU or sampler wave uses the sparse fixed-light problem. (2) Human authorisation for GPU wave 1 attempt 2 (nautilus+emcee, --array=0-11%4) after git pull of autolens_inference on the RAL login node to ca505be. (3) numba_cpu dense DROPPED by human decision 2026-10-18 (decisions/hst-cleanslate-drop-numba-dense.md): no rerun. (4) Wait for 412539, 412541 and 412565 to finish (jax_cpu sparse source_pix[1] near its end), then pull and read. (5) Then start_dev the filed Mind prompt draft/refactor/autolens_inference/flatten_cleanslate_output_layout.md (Mind 734e1bf8; human 2026-10-18: flat output/autolens/imaging/... layout, no dated folder; blocked until those arrays finish).' - recent_progress: 'Read 2026-10-18 (checkins/2026-10-18T170000Z.md). A100 base complete: 412535 (dense) and 412537 (sparse) rows for both light modes and seeds read. Dense fixed-light mass_total[1] is the best-recovered row (worst abs delta-sigma 0.24 / 0.34). Sparse fixed-light mass_total[1] diverged on both seeds (log Z -43567 / -43202, max logL ~-43500 vs ~+31649 linear; slope pinned at 2.96); observation recorded in Cortex, human reading pending. jax_cpu dense seed 1 linear row written (chain ~21 h vs ~40 min on A100; worst 1.29 sigma). numba_cpu dense 412563 FAILED at the first source_pix[1] batch (two Nautilus pool workers SIGSEGV, MaxRSS 356 MB, not OOM). Wave-1 fix merged as autolens_inference#34 (ca505be) 2026-10-17T19:25Z; attempt 2 not submitted.' - blockers: 'Sparse fixed-light mass-stage divergence unexplained (needs reproduction); wave 1 attempt 2 needs RAL checkout pull (b91ce9c -> ca505be) and human authorisation; numba_cpu dense dropped 2026-10-18 (human decision), no longer a blocker.' - job_status: 'Read 2026-10-18 ~16:55Z (sacct/squeue by the architect), pin 2026.10.17.1. 412535 (jax_gpu dense) and 412537 (jax_gpu sparse) COMPLETED 0:0 both seeds. 412539 (jax_cpu dense) RUNNING 23 h: seed 1 linear row written, seed 1 on fixed mass_total[1], seed 0 in linear mass_total[1]. 412541 (jax_cpu sparse) RUNNING 23 h, both seeds in source_pix[1], no row. 412563 (numba_cpu dense) FAILED 1:0 both seeds after 00:00:57 (pool-worker SIGSEGV; failed rows). 412565 (numba_cpu sparse) RUNNING 21 h, both seeds in source_pix[1], no row. 412543 wave 1 attempt 1: 18 arms FAILED, 18-23 cancelled; attempt 2 not submitted.' - evidence: https://github.com/PyAutoLabs/autolens_inference/pull/34 + status: needs-decision + updated: '2026-10-19' + next: 'Nothing is queued (scheduler empty after the human-directed cancel 2026-10-19 ~15:51Z). Each step needs human authorisation: (1) Confirm keeping output/autolens/prepared/ (agent proposal, not yet confirmed: wave.py needs output/autolens/prepared//seed_//manifest.json for every sampler wave and the divergence repro needs it; only re-running the base mass stage regenerates it). (2) File a Mind bug task and run a laptop repro of the jax_gpu sparse fixed-light mass_total[1] divergence (both seeds) before any CPU or sampler wave uses the sparse fixed-light problem. (3) Rerun the cancelled sparse legs under the flat layout: jax_cpu sparse seed 1 from source_pix[2], jax_cpu sparse lightfixed mass_total[1] (both seeds), numba_cpu sparse lightfixed mass_total[1] (both seeds). (4) GPU wave 1 attempt 2 (nautilus+emcee, --array=0-11%4). The RAL checkout is now at c3ca4de, which contains the #34 fix, and the attempt-1 arms are archived. (5) Commit the clean-slate rows through an autolens_inference PR. numba_cpu dense stays dropped (decisions/hst-cleanslate-drop-numba-dense.md).' + recent_progress: '2026-10-19 (checkins/2026-10-19T171637Z.md, focused clean-up): 412539 (jax_cpu dense) COMPLETED. On the human''s decision, 412541 (jax_cpu sparse) and 412565 (numba_cpu sparse) were cancelled with partial stages left on disk. jax_cpu sparse: seed 0 has slam_base complete; seed 1 has source_lp[1] and source_pix[1] complete. numba sparse: slam_base complete on both seeds. All lightfixed mass_total[1] stages are incomplete. autolens_inference#36 (flatten, closes #35) merged as c3ca4de 2026-10-19T15:53:33Z. flatten_cleanslate.py --apply ran on RAL (28 moves, 55 rewrites) and on the laptop mirror (28 moves, 38 rewrites). Stage output now lives at output/autolens/imaging/hst/slam///seed_//, prepared problems at output/autolens/prepared//seed_// and rows at results/slam/imaging/hst///. RAL and laptop each hold 1974 files with identical paths and sizes. The wave-1 attempt-1 material (output/experiments, output/experiment_sources, results/searches/experiments) was moved aside to /mnt/ral/jnightin/inference/aside/autolens_inference_cleanslate_wave1_attempt1/ on RAL and to /mnt/c/Users/Jammy/Science/inference/aside/autolens_inference_cleanslate_wave1_attempt1/ on the laptop. Earlier (2026-10-18): A100 base complete; the sparse fixed-light mass_total[1] diverged on both seeds (Cortex observation, human reading pending).' + blockers: 'Sparse fixed-light mass-stage divergence unexplained (needs reproduction). Reruns of the cancelled sparse legs and wave 1 attempt 2 need human authorisation. Keeping prepared/ awaits human confirmation. The RAL checkout pull blocker is cleared (c3ca4de).' + job_status: 'Read 2026-10-19 15:49:05Z (squeue/sacct on RAL by the architect), pin 2026.10.17.1. 412539 (jax_cpu dense) COMPLETED: 412539_0 ended 2026-10-18T23:46:01, 412539_1 ended 2026-10-18T20:43:11. 412541 (jax_cpu sparse) and 412565 (numba_cpu sparse) were RUNNING; the human had them scancelled ~15:51Z (sacct CANCELLED, elapsed 1-23:04:43 / 1-20:52:11). The queue was empty afterwards. Earlier: 412535 and 412537 (A100) COMPLETED; 412563 (numba_cpu dense) FAILED (dropped); 412543 wave 1 attempt 1 failed on every arm, final sacct not re-read. Compute submitted 2026-10-19: none.' + evidence: https://github.com/PyAutoLabs/autolens_inference/pull/36 project: autolens_inference - checkin_report: checkins/2026-10-18T170000Z.md + checkin_report: checkins/2026-10-19T171637Z.md - id: point-source-admission title: Point-source sampler admission status: needs-decision @@ -225,20 +225,20 @@ tasks: campaign: hst-imaging-mass-samplers title: 'HST imaging mass stage, clean slate: Nautilus base on A100 and jax_cpu at release 2026.10.17.1, linear and fixed lens light' - status: running + status: needs-decision priority: high path: tasks/hst_mass_base_light_modes.md - updated: '2026-10-18' - next: 'A100 rows of 412535 and 412537 (both light modes, seeds 0-1) read 2026-10-18; jax_gpu sparse fixed-light mass_total[1] diverged on both seeds (Cortex observation, human reading pending). 412539 (jax_cpu dense) has the seed-1 linear row; 412541 (jax_cpu sparse) and 412565 (numba_cpu sparse) RUNNING in source_pix[1] at ~16:55Z; 412563 (numba_cpu dense) FAILED (pool-worker SIGSEGV, not OOM), rerun with --cores 1 or NUMBA_THREADING_LAYER=workqueue needs a human decision. Pull and read 412539, 412541 and 412565 when they finish.' - evidence: https://github.com/PyAutoLabs/autolens_inference/pull/26 - checkin_report: checkins/2026-10-18T170000Z.md + updated: '2026-10-19' + next: 'A100 legs complete; jax_gpu sparse fixed-light mass_total[1] diverged (Cortex observation, human reading pending). 412539 (jax_cpu dense) COMPLETED 2026-10-18; its rows were not re-read in the 2026-10-19 record. 412541 (jax_cpu sparse) and 412565 (numba_cpu sparse) were CANCELLED 2026-10-19 ~15:51Z on the human''s decision. Completed at cancel: jax_cpu sparse seed 0 slam_base; jax_cpu sparse seed 1 source_lp[1] and source_pix[1]; numba sparse slam_base on both seeds. Every lightfixed mass_total[1] is incomplete. The partial stage dirs stay under output/autolens/imaging/hst/slam///seed_/. A rerun of the remainder under the flat layout needs human authorisation. numba_cpu dense dropped (2026-10-18 decision).' + evidence: https://github.com/PyAutoLabs/autolens_inference/pull/36 + checkin_report: checkins/2026-10-19T171637Z.md - id: hst_mass_samplers_wave1 campaign: hst-imaging-mass-samplers title: 'HST imaging mass stage: sampler comparison, wave 1' - status: blocked + status: needs-decision priority: normal path: tasks/hst_mass_samplers_wave1.md - updated: '2026-10-18' - next: 'Fix merged as autolens_inference#34 (ca505be, 2026-10-17T19:25Z): row writer keyed by dotted path, --attempt, --sampler-filter, NUTS/SMC deferred, GPU submit defaults attempt 2 nautilus+emcee --array=0-11%4 with no dependency. Attempt 2 NOT submitted; blocked on git pull of autolens_inference on the RAL login node (at b91ce9c) to ca505be and human authorisation. Attempt-1 NUTS/SMC rows exist on RAL as failed: RESOURCE_EXHAUSTED (385.32 GiB / 48.45 GiB); no nautilus/emcee rows.' + updated: '2026-10-19' + next: 'Fix merged as autolens_inference#34; the RAL checkout was fast-forwarded to c3ca4de (includes #34 and the #36 flat layout) on 2026-10-19, so the pull blocker is cleared. Attempt 2 (nautilus+emcee, --array=0-11%4, NUTS/SMC deferred) is NOT submitted and needs human authorisation. It reads output/autolens/prepared//seed_0//manifest.json. Attempt-1 material (output/experiments 341M, output/experiment_sources 76M, the 8 failed NUTS/SMC rows from results/searches/experiments) was moved aside 2026-10-19 on the human''s decision to /mnt/ral/jnightin/inference/aside/autolens_inference_cleanslate_wave1_attempt1/ (RAL) and /mnt/c/Users/Jammy/Science/inference/aside/autolens_inference_cleanslate_wave1_attempt1/ (laptop), so attempt 2 starts clean.' evidence: https://github.com/PyAutoLabs/autolens_inference/pull/34 - checkin_report: checkins/2026-10-18T170000Z.md + checkin_report: checkins/2026-10-19T171637Z.md diff --git a/checkins/2026-10-19T171637Z.md b/checkins/2026-10-19T171637Z.md new file mode 100644 index 0000000..31812ff --- /dev/null +++ b/checkins/2026-10-19T171637Z.md @@ -0,0 +1,99 @@ +# Inference check-in — 2026-10-19T17:16:37Z + +Previous: checkins/2026-10-18T170000Z.md +Scope: focused record, not a complete sweep. User focus: hst-imaging-mass-samplers clean-up (cancel the running sparse legs, flatten the clean-slate layout, set the wave-1 attempt-1 material aside). The other open campaigns were not re-swept in this session. Their rows are unchanged, and `last_checkin` / `last_checkin_report` stay at the 2026-10-18 sweep, because CHECKIN.md step 6 stamps those only for a completed sweep. +User decisions (verbatim, inference chat 2026-10-19): +- "I want us to check the on going inference campaign on RAL, but I am ready for us to update folder structures on RAL and locally and clean things up so cancel on going jobs and we'll work on the clean up" +- "yes, go" (merge of autolens_inference#36, the flatten PR) +- "check some older chats, think we dont need prepared, experiments and experiment_sources to exist anymore? Can we archive them somewhere separate to makr things easier to follow. Then do the apply". This repeats the 2026-10-18 statement: "I think all this experiment_sources, experiments is stuff i dont care to retain ... all this stuff can go somewhere safe on RAL But out of the way from our clean slate investigation". + +## Sources inspected + +| Source | Commit / query time | Evidence | Availability | +|---|---|---|---| +| RAL scheduler (`squeue`/`sacct`, euclid_jump) | 2026-10-19 15:49:05Z, plus `sacct` after the ~15:51Z cancel | queried by the session architect over SSH | verified | +| RAL stage `.completed` markers for 412541 / 412565 | 2026-10-19, after the cancel | autolens_inference RAL checkout `output/` | verified | +| autolens_inference#36 (flatten, closes #35) | merged 2026-10-19T15:53:33Z, merge commit `c3ca4de` | https://github.com/PyAutoLabs/autolens_inference/pull/36 | verified (PR state re-read 17:16Z) | +| autolens_inference issue #35 | closed | https://github.com/PyAutoLabs/autolens_inference/issues/35 | verified | +| Mind close-out | `b7cbbd58` | `complete/2026/10/flatten-cleanslate-output-layout.md` | reported by the architect; not re-read here | +| autolens_inference RAL checkout | `b91ce9c` -> `c3ca4de` (fast-forward) | RAL login node | verified by the architect | +| autolens_inference laptop checkout | `c3ca4de` on `main` | `lens/autolens_inference` | verified (re-read 17:16Z) | +| Laptop `hpc/sync pull` | completed 2026-10-19, before the migration | `.cortex/pull.json` | verified by the architect | +| `flatten_cleanslate.py --apply`, RAL and laptop | 2026-10-19 | script output, both EXIT 0 | verified by the architect | +| Post-migration listing, RAL vs laptop | 2026-10-19 | `output/autolens` + `results/slam` (search_internal excluded) | verified: 1974 files each, identical paths and sizes | +| `hpc/sync status` dry run | 2026-10-19, after the migration | 0 transferred | verified by the architect | +| Other campaigns' projects (autofit_inference, graphical_ep, analytic_gaussian, ep_toy_gaussian, slope_hierarchy_scale, ic50_workspace, point-source records) | not re-inspected | ledger state retained | not re-inspected | + +## Campaign sweep + +| Campaign | Change since previous | Coordination state | Job evidence | Blocker | Next bounded step | +|---|---|---|---|---|---| +| hst-imaging-mass-samplers | 412539 COMPLETED. The human cancelled 412541 and 412565. Flatten #36 was merged and applied on RAL and on the laptop. Wave-1 attempt-1 material set aside | needs-decision (no job queued) | squeue 2026-10-19 15:49:05Z: only 412541_0/_1 and 412565_0/_1 RUNNING. Both were scancelled ~15:51Z (sacct CANCELLED). 412539 COMPLETED. Queue empty afterwards | Sparse fixed-light divergence still unexplained. Reruns, wave 1 attempt 2 and the repro all need human authorisation | Human chooses among the proposals below; no compute is queued | +| slam-hst-base | not re-swept | superseded | not queried | Unchanged | Unchanged | +| slam-hst-delaunay | not re-swept | superseded | not queried | Unchanged | Unchanged | +| point-source-admission | not re-swept | needs-decision | not queried | Unchanged | Unchanged | +| graphical-ep | not re-swept | needs-slicing | not queried | Unchanged | Unchanged | +| analytic-gaussian | not re-swept | blocked | not queried | Unchanged | Unchanged | +| ep-toy-gaussian | not re-swept | needs-decision | not queried | Unchanged | Unchanged | +| slope-hierarchy-scale | not re-swept | blocked | not queried | Unchanged | Unchanged | +| ic50-scale | not re-swept | blocked | not queried | Unchanged | Unchanged | +| gaussian-x3-search-benchmark | not re-swept | running | not queried | Unchanged | Unchanged | + +### hst-imaging-mass-samplers job detail + +| Job | Leg | State | End / elapsed | +|---|---|---|---| +| 412539 | jax_cpu dense (ral) | COMPLETED | 412539_0 ended 2026-10-18T23:46:01, 412539_1 ended 2026-10-18T20:43:11 | +| 412541 | jax_cpu sparse (ral) | CANCELLED (scancel, human decision) ~15:51Z | elapsed 1-23:04:43 | +| 412565 | numba_cpu sparse (ral) | CANCELLED (scancel, human decision) ~15:51Z | elapsed 1-20:52:11 | + +Stages complete at cancel, from the `.completed` markers. This is execution completion only: + +| Leg / seed | slam_base complete | Incomplete (partial dirs left on disk) | +|---|---|---| +| jax_cpu sparse / 0 | light[1], source_lp[1], source_pix[1], source_pix[2], mass_total[1] | slam_base_lightfixed mass_total[1] (was running) | +| jax_cpu sparse / 1 | source_lp[1], source_pix[1] | source_pix[2] (was running), mass_total[1], whole lightfixed leg | +| numba_cpu sparse / 0 | light[1], source_pix[1], source_pix[2], mass_total[1] (source_lp shared from jax_cpu via `--stages-from`) | slam_base_lightfixed mass_total[1] (was running) | +| numba_cpu sparse / 1 | light[1], source_pix[1], source_pix[2], mass_total[1] (source_lp shared) | slam_base_lightfixed mass_total[1] (was running) | + +The incomplete stage directories are partial, not failures. + +### Layout migration (autolens_inference#36, `scripts/misc/flatten_cleanslate.py --apply`) + +- RAL: "Done: 28 move(s), 55 rewrite(s), 8 empty dir(s) removed." EXIT 0. +- Laptop mirror (`--output-root /mnt/c/Users/Jammy/Science/inference/autolens_inference/output`): "Done: 28 move(s), 38 rewrite(s), 8 empty dir(s) removed." EXIT 0. +- Stage output now lives at `output/autolens/imaging/hst/slam///seed_//`. Prepared problems are at `output/autolens/prepared//seed_//`. Result rows are at `results/slam/imaging/hst///`, with no `__cleanslate_2026_10_17` suffix. Neither side has a `cleanslate_2026_10_17` directory any more. +- Verification: RAL and the laptop each hold 1974 files under `output/autolens` + `results/slam`, with identical paths and sizes. `hpc/sync status` dry run transferred 0. + +### Set aside (wave 1 attempt 1, all arms failed) + +All attempt-1 arms failed. NUTS and SMC ran out of memory. The nautilus and emcee fits completed, but their rows were never written. The following folders were moved out of the clean-slate tree: + +- RAL: `output/experiments` (341M), `output/experiment_sources` (76M) and `results/searches/experiments` (the 8 failed NUTS/SMC row JSONs) were moved to `/mnt/ral/jnightin/inference/aside/autolens_inference_cleanslate_wave1_attempt1/{output,results/searches}/`. +- Laptop: the same folders were moved to `/mnt/c/Users/Jammy/Science/inference/aside/autolens_inference_cleanslate_wave1_attempt1/`. + +## Coverage and limits + +- `.completed` markers record execution only. They do not establish convergence or scientific acceptance. +- No CPU fixed-light row exists. Every lightfixed mass_total[1] on the CPU sparse legs was cancelled before it finished. jax_cpu sparse seed 1 is missing source_pix[2] onward. +- The 412539 (jax_cpu dense) rows were not re-read in this session. Their completion is from sacct only. +- No final sacct state for 412543 (wave 1 attempt 1) was read today. It was not in the 15:49:05Z queue. +- The clean-slate result rows are still untracked in the autolens_inference checkout. Committing them is a separate project PR and was not done here. +- The sparse fixed-light divergence observation (Cortex, 2026-10-18) still awaits the human's reading. This report draws no conclusion from it. +- The other campaigns were not re-inspected. Their rows are unchanged. + +## Decisions and handoff + +Authorised and done this session: cancellation of 412541 and 412565, the merge of autolens_inference#36, the migration `--apply` on RAL and the laptop, and the move-aside of the wave-1 attempt-1 material. All of these were on the human's words quoted above. + +Proposals (not authorised): + +1. **Keep `prepared/`.** This is the agent's recommendation. It was stated to the human but has not been explicitly confirmed. wave.py requires `output/autolens/prepared//seed_//manifest.json` for every sampler wave, and the sparse fixed-light divergence repro needs it. It can only be regenerated by re-running the base mass stage. +2. Rerun the cancelled sparse legs under the new layout: the jax_cpu sparse seed-1 remainder, the jax_cpu sparse lightfixed legs and the numba_cpu sparse lightfixed legs. +3. GPU wave 1 attempt 2 (nautilus+emcee). It now starts clean because the attempt-1 arms are archived, and the RAL checkout is at `c3ca4de`, which contains the #34 fix. +4. A Mind bug task plus a laptop repro for the jax_gpu sparse fixed-light mass_total[1] divergence. +5. Commit the clean-slate rows through an autolens_inference PR. + +Mind: flatten task completed (`b7cbbd58`, `complete/2026/10/flatten-cleanslate-output-layout.md`). No new Mind issue was opened. + +Compute submitted: none. The only scheduler action was the human-directed `scancel` of 412541 and 412565. diff --git a/dashboard.html b/dashboard.html index b036dab..749b05a 100644 --- a/dashboard.html +++ b/dashboard.html @@ -369,7 +369,7 @@ - PyAutoInsight: https://github.com/PyAutoLabs/PyAutoInsight - autolens_inference: https://github.com/PyAutoLabs/autolens_inference - autofit_inference: https://github.com/PyAutoLabs/autofit_inference
Last updated unavailable↻ Update

-

Last check-in: 2026-10-18T17:00:00Z.

Active campaigns

CampaignRecent progressNext steps
HST imaging, clean slate: Nautilus base on A100, jax_cpu and numba_cpu at a pinned release, mass stage with linear lens light and with it fixed, then other samplers on the mass stageRead 2026-10-18 (checkins/2026-10-18T170000Z.md). A100 base complete: 412535 (dense) and 412537 (sparse) rows for both light modes and seeds read. Dense fixed-light mass_total[1] is the best-recovered row (worst abs delta-sigma 0.24 / 0.34). Sparse fixed-light mass_total[1] diverged on both seeds (log Z -43567 / -43202, max logL ~-43500 vs ~+31649 linear; slope pinned at 2.96); observation recorded in Cortex, human reading pending. jax_cpu dense seed 1 linear row written (chain ~21 h vs ~40 min on A100; worst 1.29 sigma). numba_cpu dense 412563 FAILED at the first source_pix[1] batch (two Nautilus pool workers SIGSEGV, MaxRSS 356 MB, not OOM). Wave-1 fix merged as autolens_inference#34 (ca505be) 2026-10-17T19:25Z; attempt 2 not submitted.(1) Human decision on filing a Mind bug task for the jax_gpu sparse fixed-light mass_total[1] divergence (both seeds) and reproducing it on the laptop from the exported prepared problem, before any CPU or sampler wave uses the sparse fixed-light problem. (2) Human authorisation for GPU wave 1 attempt 2 (nautilus+emcee, --array=0-11%4) after git pull of autolens_inference on the RAL login node to ca505be. (3) numba_cpu dense DROPPED by human decision 2026-10-18 (decisions/hst-cleanslate-drop-numba-dense.md): no rerun. (4) Wait for 412539, 412541 and 412565 to finish (jax_cpu sparse source_pix[1] near its end), then pull and read. (5) Then start_dev the filed Mind prompt draft/refactor/autolens_inference/flatten_cleanslate_output_layout.md (Mind 734e1bf8; human 2026-10-18: flat output/autolens/imaging/... layout, no dated folder; blocked until those arrays finish).
Point-source sampler admission2 October project record reports five Nautilus seeds; no gradient-sampler admission measurement.Assess a bounded gradient-sampler admission proposal with the human; no new job is authorized.
Graphical and EP inference coordinationMind campaign map and two scoping contracts preserved verbatim; implementation work remains in Mind.Reconcile current Cortex observations and shipped fixes, then select one bounded phase.
Analytic Gaussian ensemble and control calibrationCortex records accepted wave-1 baseline and 30 September independent-review caveats.Use Cortex planned diagnostic on passing, edge and worst seeds; retain the human conclusion there.
N=50 NUTS versus three EP repeatsCortex records failed wave 1 (EMFILE/OOM), wave-2 preparation and no submission.Present the existing wave-2 proposal with current dependency evidence for decision.
Hierarchical slope recovery at scaleCortex retains graphical baseline, post-processing OOM and stale EP hierarchical factor; Mind records later moments fix.Verify the moments fix release and project mirror; recover missing summaries from existing samples before a proposed rerun.
IC50 EP and graphical scale ladderCortex records N=5/10/25 results, failed N=50 EP rung and width-interpretation caveat.Check implementation close-outs, then propose a seeded ladder with appropriate output and mean-field diagnostics.
PyAutoFit search benchmark on gaussian_x3 (autofit_inference)Wave 2 closed (B5b-2, 2026-10-15, autofit_inference#12) — RAL 409930 pulled, 12,750 rows admitted at the pin df6d218f10cd with 0 missing; NSS 50-step leads both evidence rankings; the catalogue ranks only legs with success rate >= 50 % over >= 20 attempts (post-hoc presentation convention); fit records published as inference-summary@3 (101 record shards).B5c-2b (draft/research/autofit_inference/search_ext_b5c2b_scored_wave_submit.md): realisations 2-3 and their exact references, the correlated_nd@2 scored-wave manifest (83 cells x 50 seeds x 3 realisations, 12 450 runs) and its probe for 8 probe-first cells, then the submit on RAL --partition=ral from a private pinned copy with the A2 memory requests; correlated_nd@2 is frozen (autofit_inference 44c5e77f1fb0); the family is catalogued apart from gaussian_x3 and ranks nothing across families.

Active tasks

TaskCampaignStatusPriorityNext steps
Drop the 1e-3 prior centring for shear / multipole / ell_comps in the benchmarkslam-hst-baseblockednormalVerify the release containing Galaxy#634 and Lens#754 before changing priors; retain zero-centre gradient witness. Carried into hst-imaging-mass-samplers phase 1 as a prior-identity question for the frozen model.
EP campaign — phase map for the 2026 Q3 graphical/EP pushgraphical-epneeds-decisionhighReconcile current Cortex records and human gates; this campaign map is never issued or bulk scheduled.
Expectation Propagation Scale-Up — Scopinggraphical-epneeds-slicinghighReconcile completed wrapper work and Pulse profiling ownership before selecting a bounded remaining inference phase.
Graphical Model Scale-Up — Scopinggraphical-epneeds-slicinghighReconcile current Cortex scale results and library dependencies; select one bounded inference phase with a witness.
NUTS/HMC trial on validated likelihoodspoint-source-admissionneeds-decisionnormalAn unscoped idea, not permission to submit: reconcile point-source admission with imaging and weak-lensing candidate targets.
gaussian_x3 search benchmark — wave 1 (the pilot)gaussian-x3-search-benchmarkblockednormalPilot stopped at 223 of 520 runs and is not restarted; freeze gaussian_x3@2 from its 17-row calibration before any scored run; the deferred cells move to scored wave 2 (after A2 and A4).
gaussian_x3 search benchmark — wave 2 (scored)gaussian-x3-search-benchmarkrunninghighWave 2 is closed and judged (B5b-2); the fit snapshot is captured from main under inference-summary@3 after the B5b-2 PRs merge.
HST imaging mass stage, clean slate: Nautilus base on A100 and jax_cpu at release 2026.10.17.1, linear and fixed lens lighthst-imaging-mass-samplersrunninghighA100 rows of 412535 and 412537 (both light modes, seeds 0-1) read 2026-10-18; jax_gpu sparse fixed-light mass_total[1] diverged on both seeds (Cortex observation, human reading pending). 412539 (jax_cpu dense) has the seed-1 linear row; 412541 (jax_cpu sparse) and 412565 (numba_cpu sparse) RUNNING in source_pix[1] at ~16:55Z; 412563 (numba_cpu dense) FAILED (pool-worker SIGSEGV, not OOM), rerun with --cores 1 or NUMBA_THREADING_LAYER=workqueue needs a human decision. Pull and read 412539, 412541 and 412565 when they finish.
HST imaging mass stage: sampler comparison, wave 1hst-imaging-mass-samplersblockednormalFix merged as autolens_inference#34 (ca505be, 2026-10-17T19:25Z): row writer keyed by dotted path, --attempt, --sampler-filter, NUTS/SMC deferred, GPU submit defaults attempt 2 nautilus+emcee --array=0-11%4 with no dependency. Attempt 2 NOT submitted; blocked on git pull of autolens_inference on the RAL login node (at b91ce9c) to ca505be and human authorisation. Attempt-1 NUTS/SMC rows exist on RAL as failed: RESOURCE_EXHAUSTED (385.32 GiB / 48.45 GiB); no nautilus/emcee rows.

Ledger

Inference Results

Latest refresh failed. Showing cached evidence.

PyAutoLens
Imaging
Delaunay (JAX) (opens in a new tab)Delaunay (Numba) (opens in a new tab)Rectangular (JAX) (opens in a new tab)Rectangular (Numba) (opens in a new tab)
Point Source
Simple (JAX) (opens in a new tab)
PyAutoFit

Integrity: ok · Freshness: freshness policy unspecified · evidence 2026-10-15T16:38:44Z · Scientific qualification: 5129/12973 accepted by producer; remaining assessments shown below

Project dashboard · Ingest receipt

Record shards

inference-summary@3: the records are held in one file per protocol and leg, each verified against the index's sha256 at the captured commit. Counts are the producer's assertions, shown without ranking; each shard file holds every record's diagnostics, timings and provenance.

ProtocolShardRecordsExecutionConvergenceAcceptanceEvidence
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/bfgs/default/local_jax_cpu_fp64_cache_warm10complete 2, failed 8not_assessed 10not_assessed 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/bfgs/default/local_numpy_fp6410failed 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/blackjax_nuts/warmup_200_samples_500/local_jax_cpu_fp64_cache_warm2failed 2not_converged 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/blackjax_nuts_warm/warmup_200_samples_500/local_jax_cpu_fp64_cache_warm1failed 1not_converged 1not_assessed 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/dynesty_dynamic/default/local_jax_cpu_fp64_cache_warm2complete 2not_assessed 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/dynesty_dynamic/default/local_numpy_fp642complete 2not_assessed 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/dynesty_static/nlive_200/local_jax_cpu_fp64_cache_warm2complete 2not_assessed 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/dynesty_static/nlive_200/local_numpy_fp642complete 2not_assessed 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/emcee/walkers_50_steps_2000/local_jax_cpu_fp64_cache_warm2complete 1, failed 1not_converged 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/emcee/walkers_50_steps_2000/local_numpy_fp642complete 2not_converged 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/lbfgs/default/local_jax_cpu_fp64_cache_warm10complete 1, failed 9not_assessed 10not_assessed 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/lbfgs/default/local_numpy_fp6410complete 2, failed 8not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/multi_start_adabelief/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10not_assessed 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/multi_start_adam/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10not_assessed 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/multi_start_lion/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10not_assessed 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/multi_start_prodigy/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10not_assessed 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/nautilus/n_live_100/local_jax_cpu_fp64_cache_warm2complete 2converged 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/nautilus/n_live_100/local_numpy_fp642complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/nautilus/n_live_200/local_jax_cpu_fp64_cache_warm2complete 2converged 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/nautilus/n_live_200/local_numpy_fp642complete 1, failed 1converged 1, not_converged 1accepted 1, rejected 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/nautilus/n_live_400/local_jax_cpu_fp64_cache_warm2complete 2converged 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/nautilus/n_live_400/local_numpy_fp642complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/smc/default/local_jax_cpu_fp64_cache_warm2complete 2not_assessed 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/zeus/walkers_50_steps_2000/local_jax_cpu_fp64_cache_warm1failed 1not_converged 1not_assessed 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/zeus/walkers_50_steps_2000/local_numpy_fp642complete 1, failed 1not_converged 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/bfgs/default/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/bfgs/default/local_numpy_fp6410complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/blackjax_nuts/warmup_200_samples_500/local_jax_cpu_fp64_cache_warm2complete 2not_converged 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/blackjax_nuts_warm/warmup_200_samples_500/local_jax_cpu_fp64_cache_warm1complete 1not_converged 1accepted 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/dynesty_dynamic/default/local_jax_cpu_fp64_cache_warm2complete 2not_assessed 2accepted 1, rejected 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/dynesty_dynamic/default/local_numpy_fp642complete 2not_assessed 2accepted 1, rejected 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/dynesty_static/nlive_200/local_jax_cpu_fp64_cache_warm2complete 2not_assessed 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/dynesty_static/nlive_200/local_numpy_fp642complete 2not_assessed 2accepted 1, rejected 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/emcee/walkers_50_steps_2000/local_jax_cpu_fp64_cache_warm2complete 2not_converged 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/emcee/walkers_50_steps_2000/local_numpy_fp642complete 2not_converged 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/lbfgs/default/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/lbfgs/default/local_numpy_fp6410complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/multi_start_adabelief/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/multi_start_adam/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/multi_start_lion/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/multi_start_prodigy/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10accepted 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/nautilus/n_live_100/local_jax_cpu_fp64_cache_warm2complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/nautilus/n_live_100/local_numpy_fp642complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/nautilus/n_live_200/local_jax_cpu_fp64_cache_warm2complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/nautilus/n_live_200/local_numpy_fp642complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/nautilus/n_live_400/local_jax_cpu_fp64_cache_warm2complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/nautilus/n_live_400/local_numpy_fp642complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/smc/default/local_jax_cpu_fp64_cache_warm1complete 1not_assessed 1rejected 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/zeus/walkers_50_steps_2000/local_jax_cpu_fp64_cache_warm1failed 1not_converged 1rejected 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/zeus/walkers_50_steps_2000/local_numpy_fp642complete 2not_converged 2accepted 2records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/bfgs/default/ral_jax_cpu_fp64_cache_warm250failed 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/bfgs/default/ral_numpy_fp64250failed 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/blackjax_nuts/warmup_200_samples_500/ral_jax_cpu_fp64_cache_warm250failed 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/dynesty_dynamic/default/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 59, rejected 191records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/dynesty_dynamic/default/ral_numpy_fp64250complete 250not_assessed 250accepted 59, rejected 191records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/dynesty_static/nlive_200/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 36, rejected 214records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/dynesty_static/nlive_200/ral_numpy_fp64250complete 250not_assessed 250accepted 36, rejected 214records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/emcee/walkers_50_steps_2000/ral_jax_cpu_fp64_cache_warm250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/emcee/walkers_50_steps_2000/ral_numpy_fp64250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/lbfgs/default/ral_jax_cpu_fp64_cache_warm250complete 40, failed 210not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/lbfgs/default/ral_numpy_fp64250complete 40, failed 210not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/multi_start_adabelief/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/multi_start_adam/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/multi_start_lion/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/multi_start_prodigy/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 152, rejected 98records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nautilus/n_live_100/ral_jax_cpu_fp64_cache_warm250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nautilus/n_live_100/ral_numpy_fp64250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nautilus/n_live_200/ral_jax_cpu_fp64_cache_warm250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nautilus/n_live_200/ral_numpy_fp64250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nautilus/n_live_400/ral_jax_cpu_fp64_cache_warm250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nautilus/n_live_400/ral_numpy_fp64250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nss/n_live_200_mcmc_steps_50/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 158, rejected 92records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/smc/default/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/zeus/walkers_50_steps_2000/ral_jax_cpu_fp64_cache_warm250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/zeus/walkers_50_steps_2000/ral_numpy_fp64250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/bfgs/default/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/bfgs/default/ral_numpy_fp64250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/blackjax_nuts/warmup_200_samples_500/ral_jax_cpu_fp64_cache_warm250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/blackjax_nuts_warm/warmup_200_samples_500/ral_jax_cpu_fp64_cache_warm250complete 250converged 21, not_converged 229accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/dynesty_dynamic/default/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 145, rejected 105records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/dynesty_dynamic/default/ral_numpy_fp64250complete 250not_assessed 250accepted 145, rejected 105records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/dynesty_static/nlive_200/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 103, rejected 147records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/dynesty_static/nlive_200/ral_numpy_fp64250complete 250not_assessed 250accepted 103, rejected 147records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/emcee/walkers_50_steps_2000/ral_jax_cpu_fp64_cache_warm250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/emcee/walkers_50_steps_2000/ral_numpy_fp64250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/lbfgs/default/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/lbfgs/default/ral_numpy_fp64250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/multi_start_adabelief/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/multi_start_adam/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/multi_start_lion/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/multi_start_prodigy/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nautilus/n_live_100/ral_jax_cpu_fp64_cache_warm250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nautilus/n_live_100/ral_numpy_fp64250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nautilus/n_live_200/ral_jax_cpu_fp64_cache_warm250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nautilus/n_live_200/ral_numpy_fp64250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nautilus/n_live_400/ral_jax_cpu_fp64_cache_warm250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nautilus/n_live_400/ral_numpy_fp64250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nss/n_live_200_mcmc_steps_50/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 210, rejected 40records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/smc/default/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/zeus/walkers_50_steps_2000/ral_jax_cpu_fp64_cache_warm250complete 250not_converged 250accepted 190, rejected 60records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/zeus/walkers_50_steps_2000/ral_numpy_fp64250complete 250not_converged 250accepted 198, rejected 52records

Comparisons and parity groups

Targets, seeds and stages are retained. No best-seed selection, cross-target ranking or automatic convergence threshold. Declared groups do not themselves certify parity.

No certified comparison pairs supplied by this producer.

Decision History

Drop the numba_cpu dense leg from the HST clean-slate campaign

Sampler candidates

Literature expectations, not measured PyAuto recommendations. Review a candidate and copy its investigation prompt.

dynesty

dynamic nested sampling · reviewed 2026-10-08

Paper ↗ · Official code ↗

Expected uses

  • Evidence and posterior baseline with adjustable effort allocation.

Limitations

  • Prior transforms and likelihood-contour exploration need validation; no universal dimensional scaling advantage.

Integration: integrated: DynestyStatic and DynestyDynamic

Benchmark evidence: not_assessed: this literature review establishes no benchmark result

Investigation prompt
Nautilus

neural-network importance nested sampling · reviewed 2026-10-08

Paper ↗ · Official code ↗

Expected uses

  • Evidence and posterior estimation for expensive gradient-free likelihoods.

Limitations

  • Learned proposal coverage and importance-weight stability need checking on unseen modes.

Integration: integrated: Nautilus

Benchmark evidence: not_assessed: this literature review establishes no benchmark result

Investigation prompt
Last updated unavailable↻ Update

-

Last check-in: 2026-10-18T17:00:00Z.

Active campaigns

CampaignRecent progressNext steps
HST imaging, clean slate: Nautilus base on A100, jax_cpu and numba_cpu at a pinned release, mass stage with linear lens light and with it fixed, then other samplers on the mass stage2026-10-19 (checkins/2026-10-19T171637Z.md, focused clean-up): 412539 (jax_cpu dense) COMPLETED. On the human's decision, 412541 (jax_cpu sparse) and 412565 (numba_cpu sparse) were cancelled with partial stages left on disk. jax_cpu sparse: seed 0 has slam_base complete; seed 1 has source_lp[1] and source_pix[1] complete. numba sparse: slam_base complete on both seeds. All lightfixed mass_total[1] stages are incomplete. autolens_inference#36 (flatten, closes #35) merged as c3ca4de 2026-10-19T15:53:33Z. flatten_cleanslate.py --apply ran on RAL (28 moves, 55 rewrites) and on the laptop mirror (28 moves, 38 rewrites). Stage output now lives at output/autolens/imaging/hst/slam/<variant>/<config>/seed_<n>/<stage>/, prepared problems at output/autolens/prepared/<config>/seed_<n>/<variant>/ and rows at results/slam/imaging/hst/<variant>/<config>/. RAL and laptop each hold 1974 files with identical paths and sizes. The wave-1 attempt-1 material (output/experiments, output/experiment_sources, results/searches/experiments) was moved aside to /mnt/ral/jnightin/inference/aside/autolens_inference_cleanslate_wave1_attempt1/ on RAL and to /mnt/c/Users/Jammy/Science/inference/aside/autolens_inference_cleanslate_wave1_attempt1/ on the laptop. Earlier (2026-10-18): A100 base complete; the sparse fixed-light mass_total[1] diverged on both seeds (Cortex observation, human reading pending).Nothing is queued (scheduler empty after the human-directed cancel 2026-10-19 ~15:51Z). Each step needs human authorisation: (1) Confirm keeping output/autolens/prepared/ (agent proposal, not yet confirmed: wave.py needs output/autolens/prepared/<config>/seed_<n>/<variant>/manifest.json for every sampler wave and the divergence repro needs it; only re-running the base mass stage regenerates it). (2) File a Mind bug task and run a laptop repro of the jax_gpu sparse fixed-light mass_total[1] divergence (both seeds) before any CPU or sampler wave uses the sparse fixed-light problem. (3) Rerun the cancelled sparse legs under the flat layout: jax_cpu sparse seed 1 from source_pix[2], jax_cpu sparse lightfixed mass_total[1] (both seeds), numba_cpu sparse lightfixed mass_total[1] (both seeds). (4) GPU wave 1 attempt 2 (nautilus+emcee, --array=0-11%4). The RAL checkout is now at c3ca4de, which contains the #34 fix, and the attempt-1 arms are archived. (5) Commit the clean-slate rows through an autolens_inference PR. numba_cpu dense stays dropped (decisions/hst-cleanslate-drop-numba-dense.md).
Point-source sampler admission2 October project record reports five Nautilus seeds; no gradient-sampler admission measurement.Assess a bounded gradient-sampler admission proposal with the human; no new job is authorized.
Graphical and EP inference coordinationMind campaign map and two scoping contracts preserved verbatim; implementation work remains in Mind.Reconcile current Cortex observations and shipped fixes, then select one bounded phase.
Analytic Gaussian ensemble and control calibrationCortex records accepted wave-1 baseline and 30 September independent-review caveats.Use Cortex planned diagnostic on passing, edge and worst seeds; retain the human conclusion there.
N=50 NUTS versus three EP repeatsCortex records failed wave 1 (EMFILE/OOM), wave-2 preparation and no submission.Present the existing wave-2 proposal with current dependency evidence for decision.
Hierarchical slope recovery at scaleCortex retains graphical baseline, post-processing OOM and stale EP hierarchical factor; Mind records later moments fix.Verify the moments fix release and project mirror; recover missing summaries from existing samples before a proposed rerun.
IC50 EP and graphical scale ladderCortex records N=5/10/25 results, failed N=50 EP rung and width-interpretation caveat.Check implementation close-outs, then propose a seeded ladder with appropriate output and mean-field diagnostics.
PyAutoFit search benchmark on gaussian_x3 (autofit_inference)Wave 2 closed (B5b-2, 2026-10-15, autofit_inference#12) — RAL 409930 pulled, 12,750 rows admitted at the pin df6d218f10cd with 0 missing; NSS 50-step leads both evidence rankings; the catalogue ranks only legs with success rate >= 50 % over >= 20 attempts (post-hoc presentation convention); fit records published as inference-summary@3 (101 record shards).B5c-2b (draft/research/autofit_inference/search_ext_b5c2b_scored_wave_submit.md): realisations 2-3 and their exact references, the correlated_nd@2 scored-wave manifest (83 cells x 50 seeds x 3 realisations, 12 450 runs) and its probe for 8 probe-first cells, then the submit on RAL --partition=ral from a private pinned copy with the A2 memory requests; correlated_nd@2 is frozen (autofit_inference 44c5e77f1fb0); the family is catalogued apart from gaussian_x3 and ranks nothing across families.

Active tasks

TaskCampaignStatusPriorityNext steps
Drop the 1e-3 prior centring for shear / multipole / ell_comps in the benchmarkslam-hst-baseblockednormalVerify the release containing Galaxy#634 and Lens#754 before changing priors; retain zero-centre gradient witness. Carried into hst-imaging-mass-samplers phase 1 as a prior-identity question for the frozen model.
EP campaign — phase map for the 2026 Q3 graphical/EP pushgraphical-epneeds-decisionhighReconcile current Cortex records and human gates; this campaign map is never issued or bulk scheduled.
Expectation Propagation Scale-Up — Scopinggraphical-epneeds-slicinghighReconcile completed wrapper work and Pulse profiling ownership before selecting a bounded remaining inference phase.
Graphical Model Scale-Up — Scopinggraphical-epneeds-slicinghighReconcile current Cortex scale results and library dependencies; select one bounded inference phase with a witness.
NUTS/HMC trial on validated likelihoodspoint-source-admissionneeds-decisionnormalAn unscoped idea, not permission to submit: reconcile point-source admission with imaging and weak-lensing candidate targets.
gaussian_x3 search benchmark — wave 1 (the pilot)gaussian-x3-search-benchmarkblockednormalPilot stopped at 223 of 520 runs and is not restarted; freeze gaussian_x3@2 from its 17-row calibration before any scored run; the deferred cells move to scored wave 2 (after A2 and A4).
gaussian_x3 search benchmark — wave 2 (scored)gaussian-x3-search-benchmarkrunninghighWave 2 is closed and judged (B5b-2); the fit snapshot is captured from main under inference-summary@3 after the B5b-2 PRs merge.
HST imaging mass stage, clean slate: Nautilus base on A100 and jax_cpu at release 2026.10.17.1, linear and fixed lens lighthst-imaging-mass-samplersneeds-decisionhighA100 legs complete; jax_gpu sparse fixed-light mass_total[1] diverged (Cortex observation, human reading pending). 412539 (jax_cpu dense) COMPLETED 2026-10-18; its rows were not re-read in the 2026-10-19 record. 412541 (jax_cpu sparse) and 412565 (numba_cpu sparse) were CANCELLED 2026-10-19 ~15:51Z on the human's decision. Completed at cancel: jax_cpu sparse seed 0 slam_base; jax_cpu sparse seed 1 source_lp[1] and source_pix[1]; numba sparse slam_base on both seeds. Every lightfixed mass_total[1] is incomplete. The partial stage dirs stay under output/autolens/imaging/hst/slam/<variant>/<config>/seed_<n>/. A rerun of the remainder under the flat layout needs human authorisation. numba_cpu dense dropped (2026-10-18 decision).
HST imaging mass stage: sampler comparison, wave 1hst-imaging-mass-samplersneeds-decisionnormalFix merged as autolens_inference#34; the RAL checkout was fast-forwarded to c3ca4de (includes #34 and the #36 flat layout) on 2026-10-19, so the pull blocker is cleared. Attempt 2 (nautilus+emcee, --array=0-11%4, NUTS/SMC deferred) is NOT submitted and needs human authorisation. It reads output/autolens/prepared/<config>/seed_0/<variant>/manifest.json. Attempt-1 material (output/experiments 341M, output/experiment_sources 76M, the 8 failed NUTS/SMC rows from results/searches/experiments) was moved aside 2026-10-19 on the human's decision to /mnt/ral/jnightin/inference/aside/autolens_inference_cleanslate_wave1_attempt1/ (RAL) and /mnt/c/Users/Jammy/Science/inference/aside/autolens_inference_cleanslate_wave1_attempt1/ (laptop), so attempt 2 starts clean.

Ledger

Inference Results

Latest refresh failed. Showing cached evidence.

PyAutoLens
Imaging
Delaunay (JAX) (opens in a new tab)Delaunay (Numba) (opens in a new tab)Rectangular (JAX) (opens in a new tab)Rectangular (Numba) (opens in a new tab)
Point Source
Simple (JAX) (opens in a new tab)
PyAutoFit

Integrity: ok · Freshness: freshness policy unspecified · evidence 2026-10-15T16:38:44Z · Scientific qualification: 5129/12973 accepted by producer; remaining assessments shown below

Project dashboard · Ingest receipt

Record shards

inference-summary@3: the records are held in one file per protocol and leg, each verified against the index's sha256 at the captured commit. Counts are the producer's assertions, shown without ranking; each shard file holds every record's diagnostics, timings and provenance.

ProtocolShardRecordsExecutionConvergenceAcceptanceEvidence
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/bfgs/default/local_jax_cpu_fp64_cache_warm10complete 2, failed 8not_assessed 10not_assessed 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/bfgs/default/local_numpy_fp6410failed 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/blackjax_nuts/warmup_200_samples_500/local_jax_cpu_fp64_cache_warm2failed 2not_converged 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/blackjax_nuts_warm/warmup_200_samples_500/local_jax_cpu_fp64_cache_warm1failed 1not_converged 1not_assessed 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/dynesty_dynamic/default/local_jax_cpu_fp64_cache_warm2complete 2not_assessed 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/dynesty_dynamic/default/local_numpy_fp642complete 2not_assessed 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/dynesty_static/nlive_200/local_jax_cpu_fp64_cache_warm2complete 2not_assessed 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/dynesty_static/nlive_200/local_numpy_fp642complete 2not_assessed 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/emcee/walkers_50_steps_2000/local_jax_cpu_fp64_cache_warm2complete 1, failed 1not_converged 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/emcee/walkers_50_steps_2000/local_numpy_fp642complete 2not_converged 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/lbfgs/default/local_jax_cpu_fp64_cache_warm10complete 1, failed 9not_assessed 10not_assessed 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/lbfgs/default/local_numpy_fp6410complete 2, failed 8not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/multi_start_adabelief/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10not_assessed 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/multi_start_adam/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10not_assessed 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/multi_start_lion/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10not_assessed 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/multi_start_prodigy/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10not_assessed 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/nautilus/n_live_100/local_jax_cpu_fp64_cache_warm2complete 2converged 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/nautilus/n_live_100/local_numpy_fp642complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/nautilus/n_live_200/local_jax_cpu_fp64_cache_warm2complete 2converged 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/nautilus/n_live_200/local_numpy_fp642complete 1, failed 1converged 1, not_converged 1accepted 1, rejected 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/nautilus/n_live_400/local_jax_cpu_fp64_cache_warm2complete 2converged 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/nautilus/n_live_400/local_numpy_fp642complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/smc/default/local_jax_cpu_fp64_cache_warm2complete 2not_assessed 2not_assessed 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/zeus/walkers_50_steps_2000/local_jax_cpu_fp64_cache_warm1failed 1not_converged 1not_assessed 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_blend/zeus/walkers_50_steps_2000/local_numpy_fp642complete 1, failed 1not_converged 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/bfgs/default/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/bfgs/default/local_numpy_fp6410complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/blackjax_nuts/warmup_200_samples_500/local_jax_cpu_fp64_cache_warm2complete 2not_converged 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/blackjax_nuts_warm/warmup_200_samples_500/local_jax_cpu_fp64_cache_warm1complete 1not_converged 1accepted 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/dynesty_dynamic/default/local_jax_cpu_fp64_cache_warm2complete 2not_assessed 2accepted 1, rejected 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/dynesty_dynamic/default/local_numpy_fp642complete 2not_assessed 2accepted 1, rejected 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/dynesty_static/nlive_200/local_jax_cpu_fp64_cache_warm2complete 2not_assessed 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/dynesty_static/nlive_200/local_numpy_fp642complete 2not_assessed 2accepted 1, rejected 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/emcee/walkers_50_steps_2000/local_jax_cpu_fp64_cache_warm2complete 2not_converged 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/emcee/walkers_50_steps_2000/local_numpy_fp642complete 2not_converged 2rejected 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/lbfgs/default/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/lbfgs/default/local_numpy_fp6410complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/multi_start_adabelief/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/multi_start_adam/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/multi_start_lion/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10rejected 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/multi_start_prodigy/starts_48_steps_300/local_jax_cpu_fp64_cache_warm10complete 10not_assessed 10accepted 10records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/nautilus/n_live_100/local_jax_cpu_fp64_cache_warm2complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/nautilus/n_live_100/local_numpy_fp642complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/nautilus/n_live_200/local_jax_cpu_fp64_cache_warm2complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/nautilus/n_live_200/local_numpy_fp642complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/nautilus/n_live_400/local_jax_cpu_fp64_cache_warm2complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/nautilus/n_live_400/local_numpy_fp642complete 2converged 2accepted 2records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/smc/default/local_jax_cpu_fp64_cache_warm1complete 1not_assessed 1rejected 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/zeus/walkers_50_steps_2000/local_jax_cpu_fp64_cache_warm1failed 1not_converged 1rejected 1records
gaussian_x3@1gaussian_x3@1/gaussian_x3_separated/zeus/walkers_50_steps_2000/local_numpy_fp642complete 2not_converged 2accepted 2records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/bfgs/default/ral_jax_cpu_fp64_cache_warm250failed 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/bfgs/default/ral_numpy_fp64250failed 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/blackjax_nuts/warmup_200_samples_500/ral_jax_cpu_fp64_cache_warm250failed 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/dynesty_dynamic/default/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 59, rejected 191records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/dynesty_dynamic/default/ral_numpy_fp64250complete 250not_assessed 250accepted 59, rejected 191records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/dynesty_static/nlive_200/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 36, rejected 214records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/dynesty_static/nlive_200/ral_numpy_fp64250complete 250not_assessed 250accepted 36, rejected 214records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/emcee/walkers_50_steps_2000/ral_jax_cpu_fp64_cache_warm250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/emcee/walkers_50_steps_2000/ral_numpy_fp64250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/lbfgs/default/ral_jax_cpu_fp64_cache_warm250complete 40, failed 210not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/lbfgs/default/ral_numpy_fp64250complete 40, failed 210not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/multi_start_adabelief/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/multi_start_adam/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/multi_start_lion/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/multi_start_prodigy/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 152, rejected 98records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nautilus/n_live_100/ral_jax_cpu_fp64_cache_warm250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nautilus/n_live_100/ral_numpy_fp64250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nautilus/n_live_200/ral_jax_cpu_fp64_cache_warm250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nautilus/n_live_200/ral_numpy_fp64250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nautilus/n_live_400/ral_jax_cpu_fp64_cache_warm250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nautilus/n_live_400/ral_numpy_fp64250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/nss/n_live_200_mcmc_steps_50/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 158, rejected 92records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/smc/default/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/zeus/walkers_50_steps_2000/ral_jax_cpu_fp64_cache_warm250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_blend/zeus/walkers_50_steps_2000/ral_numpy_fp64250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/bfgs/default/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/bfgs/default/ral_numpy_fp64250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/blackjax_nuts/warmup_200_samples_500/ral_jax_cpu_fp64_cache_warm250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/blackjax_nuts_warm/warmup_200_samples_500/ral_jax_cpu_fp64_cache_warm250complete 250converged 21, not_converged 229accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/dynesty_dynamic/default/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 145, rejected 105records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/dynesty_dynamic/default/ral_numpy_fp64250complete 250not_assessed 250accepted 145, rejected 105records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/dynesty_static/nlive_200/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 103, rejected 147records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/dynesty_static/nlive_200/ral_numpy_fp64250complete 250not_assessed 250accepted 103, rejected 147records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/emcee/walkers_50_steps_2000/ral_jax_cpu_fp64_cache_warm250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/emcee/walkers_50_steps_2000/ral_numpy_fp64250complete 250not_converged 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/lbfgs/default/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/lbfgs/default/ral_numpy_fp64250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/multi_start_adabelief/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/multi_start_adam/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/multi_start_lion/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/multi_start_prodigy/starts_48_steps_300/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nautilus/n_live_100/ral_jax_cpu_fp64_cache_warm250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nautilus/n_live_100/ral_numpy_fp64250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nautilus/n_live_200/ral_jax_cpu_fp64_cache_warm250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nautilus/n_live_200/ral_numpy_fp64250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nautilus/n_live_400/ral_jax_cpu_fp64_cache_warm250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nautilus/n_live_400/ral_numpy_fp64250complete 250converged 250accepted 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/nss/n_live_200_mcmc_steps_50/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250accepted 210, rejected 40records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/smc/default/ral_jax_cpu_fp64_cache_warm250complete 250not_assessed 250rejected 250records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/zeus/walkers_50_steps_2000/ral_jax_cpu_fp64_cache_warm250complete 250not_converged 250accepted 190, rejected 60records
gaussian_x3@2gaussian_x3@2/gaussian_x3_separated/zeus/walkers_50_steps_2000/ral_numpy_fp64250complete 250not_converged 250accepted 198, rejected 52records

Comparisons and parity groups

Targets, seeds and stages are retained. No best-seed selection, cross-target ranking or automatic convergence threshold. Declared groups do not themselves certify parity.

No certified comparison pairs supplied by this producer.

Decision History

Drop the numba_cpu dense leg from the HST clean-slate campaign

Sampler candidates

Literature expectations, not measured PyAuto recommendations. Review a candidate and copy its investigation prompt.

dynesty

dynamic nested sampling · reviewed 2026-10-08

Paper ↗ · Official code ↗

Expected uses

  • Evidence and posterior baseline with adjustable effort allocation.

Limitations

  • Prior transforms and likelihood-contour exploration need validation; no universal dimensional scaling advantage.

Integration: integrated: DynestyStatic and DynestyDynamic

Benchmark evidence: not_assessed: this literature review establishes no benchmark result

Investigation prompt
Nautilus

neural-network importance nested sampling · reviewed 2026-10-08

Paper ↗ · Official code ↗

Expected uses

  • Evidence and posterior estimation for expensive gradient-free likelihoods.

Limitations

  • Learned proposal coverage and importance-weight stability need checking on unseen modes.

Integration: integrated: Nautilus

Benchmark evidence: not_assessed: this literature review establishes no benchmark result

Investigation prompt