diff --git a/campaigns.yaml b/campaigns.yaml index 3ea0692..ec0606c 100644 --- a/campaigns.yaml +++ b/campaigns.yaml @@ -1,29 +1,57 @@ version: 1 -last_checkin: null +last_checkin: '2026-10-17T11:18:53Z' +last_checkin_report: checkins/2026-10-17T111853Z.md campaigns: - id: slam-hst-base title: HST SLaM base backend parity - status: active - updated: '2026-10-04' - next: Reconcile all seeds and stages with current project rows and Cortex before proposing another run. - recent_progress: Project and Cortex ledgers retain the 24 September rerun submissions; published rows require - coverage reconciliation. - blockers: Live RAL status and output/archive availability unverified. - job_status: unknown — live scheduler not queried; historical ledger states are not current job evidence - evidence: https://github.com/PyAutoLabs/PyAutoCortex/blob/d45f198804615c090e6b6694cf725ddb0be2354e/projects/autolens_inference.md + status: superseded + updated: '2026-10-17' + next: 'Superseded 2026-10-17 by hst-imaging-mass-samplers (human: clean slate). The 2026-09-24 re-run rows + stay as history in autolens_inference PR #23; no further run under these campaigns.' + recent_progress: 'The 2026-09-24 slam_base re-runs landed as 6 rows (A100 jax_gpu dense and sparse, numba_cpu + sparse; seeds 0-1) in autolens_inference PR #23, pending merge. A100 walls are clean single-attempt; the numba_cpu + sparse rows carry a resumed source_pix[1] stage.' + blockers: The three CPU legs jax_cpu dense, jax_cpu sparse and numba_cpu dense have not run; numba_cpu sparse + chain walls are multi-attempt. + job_status: 'All 2026-09-24 re-runs COMPLETED 0:0 per sacct queried 2026-10-17T11:10Z: 350674 and 350675 (gpu), + 350682 (ral).' + evidence: https://github.com/PyAutoLabs/autolens_inference/pull/23 project: autolens_inference + checkin_report: checkins/2026-10-17T111853Z.md - id: slam-hst-delaunay title: HST SLaM Delaunay variant - status: active - updated: '2026-10-04' - next: Retrieve or account for every rerun, failed seed and missing stage; preserve the archived comparison - context. - recent_progress: Delaunay is a separate target/recipe; pre-speedup runs are archived, reruns recorded on - 24 September. - blockers: Do not merge the Delaunay target with rectangular comparisons; current job status unknown. - job_status: unknown — live scheduler not queried; historical ledger states are not current job evidence - evidence: https://github.com/PyAutoLabs/autolens_inference/blob/146a84346eaa51eda158634cae520d677f3e6213/wiki/project/state.md + status: superseded + updated: '2026-10-17' + next: 'Superseded 2026-10-17 by hst-imaging-mass-samplers (human: clean slate). The 2026-09-24 re-run rows + stay as history in autolens_inference PR #23; no further run under these campaigns.' + recent_progress: 'The 2026-09-24 delaunay_1250 re-runs landed as 6 rows (A100 jax_gpu dense and sparse, numba_cpu + sparse; seeds 0-1) in autolens_inference PR #23, pending merge. A100 walls are clean single-attempt; the numba_cpu + sparse rows carry a resumed mass_total[1] stage.' + blockers: The seed-0 mass_total[1] offset (5.98, 5.12 and 6.62 sigma on dense, sparse and numba_cpu) reproduces + on three backends; human reading pending in Cortex. Do not merge the Delaunay target with rectangular comparisons. + job_status: 'All 2026-09-24 re-runs COMPLETED 0:0 per sacct queried 2026-10-17T11:10Z: 350678 and 350679 (gpu), + 350684 (ral).' + evidence: https://github.com/PyAutoLabs/autolens_inference/pull/23 project: autolens_inference + checkin_report: checkins/2026-10-17T111853Z.md +- id: hst-imaging-mass-samplers + title: 'HST imaging, clean slate: Nautilus base on A100 and jax_cpu at a pinned release, mass stage with linear + lens light and with it fixed, then other samplers on the mass stage' + status: needs-slicing + updated: '2026-10-17' + next: 'Phase 1 via Mind start_dev: lens-light mode as a run variant (--lens-light linear|fixed|both, fixed = ordinary + profiles with intensities from the light[1] fit), both rows from one chain, --export-prepared for both modes, + library revisions and SLURM job id recorded in every row, submit scripts for A100 dense/sparse and ral jax_cpu + dense/sparse against a private pinned checkout; then phase 2 compute by separate authorisation.' + recent_progress: 'Opened 2026-10-17 as a fresh start by human decision: the base is re-measured from scratch on + PyAuto release 2026.10.17.1 (PyAutoLens tag v2026.10.17.1, released 2026-10-17) in fresh output trees; the + 2026-09-24 rows and both earlier campaigns are superseded history.' + blockers: Phase 1 runner change not started (Mind start_dev pending); no pinned private checkout exists on RAL + yet; compute not authorised until phase 1 ships. + job_status: none submitted + evidence: https://github.com/PyAutoLabs/autolens_inference/pull/23 + project: autolens_inference + checkin_report: checkins/2026-10-17T111853Z.md - id: point-source-admission title: Point-source sampler admission status: needs-decision @@ -116,7 +144,7 @@ tasks: path: tasks/drop_1e3_prior_centring_after_631.md updated: '2026-10-04' next: Verify the release containing Galaxy#634 and Lens#754 before changing priors; retain zero-centre gradient - witness. + witness. Carried into hst-imaging-mass-samplers phase 1 as a prior-identity question for the frozen model. evidence: https://github.com/PyAutoLabs/PyAutoMind/blob/8dd1edc7e04c960daae3663513c61395823eddb0/draft/feature/autolens_inference/drop_1e3_prior_centring_after_631.md - id: reconcile_cortex_ledger_references campaign: slam-hst-base @@ -199,3 +227,26 @@ tasks: next: Wave 2 is closed and judged (B5b-2); the fit snapshot is captured from main under inference-summary@3 after the B5b-2 PRs merge. evidence: https://github.com/PyAutoLabs/autofit_inference/issues/12 +- id: hst_mass_base_light_modes + 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: needs-slicing + priority: high + path: tasks/hst_mass_base_light_modes.md + updated: '2026-10-17' + next: 'Phase 1 via Mind start_dev: --lens-light linear|fixed|both, both rows from one chain, --export-prepared + for both modes, library revisions and SLURM job id in every row, A100 and ral jax_cpu submit scripts against + a private pinned checkout; no compute until phase 1 ships and phase 2 is authorised.' + evidence: https://github.com/PyAutoLabs/autolens_inference/pull/23 +- id: hst_mass_samplers_wave1 + campaign: hst-imaging-mass-samplers + title: 'HST imaging mass stage: sampler comparison, wave 1' + status: needs-slicing + priority: normal + path: tasks/hst_mass_samplers_wave1.md + updated: '2026-10-17' + next: Depends on the frozen mass_total[1] problems (both light modes, A100 and jax_cpu) from the pinned base + in hst_mass_base_light_modes; candidate samplers from the experiments runner (nautilus, emcee, nuts, smc) and + the promoted PyAutoFit catalogue (NSS, DynestyStatic/Dynamic, Zeus) are not yet sliced. + evidence: https://github.com/PyAutoLabs/autolens_inference/pull/23 diff --git a/checkins/2026-10-17T111853Z.md b/checkins/2026-10-17T111853Z.md new file mode 100644 index 0000000..9c1be9b --- /dev/null +++ b/checkins/2026-10-17T111853Z.md @@ -0,0 +1,94 @@ +# Inference check-in — 2026-10-17T11:18:53Z + +Previous: none (`last_checkin` was null; the initial migration was not a check-in) +Scope: all open campaigns; optional user focus "i want to begin our first proper Insight campaign, using PyAutoLens imaging likelihood function, starting with nautilus base for cpu and gpu (a100 on RAL) with the mass phase doing both with linear light profiles and with them removed. Get the base up and then we try other samplers on mass" +User decisions: the direction quoted above, given in the Insight chat on 2026-10-17. It opens a campaign; it authorises no compute. Later in the same chat on 2026-10-17 the human decided: "Let's do this as a fresh start we did kind of start a campaign but weve done lots of improvements since warranting a clean slate." Then, answering four explicit questions: (1) clean slate: "Yes, supersede and re-run on the pin"; (2) CPU backend: "jax_cpu, dense + sparse"; (3) light mode: "Fix lens light to ordinary profiles"; (4) test fix: "Yes, relax the two assertions". These decisions authorise no compute. + +## Sources inspected + +| Source | Commit / query time | Evidence | Availability | +|---|---|---|---| +| autolens_inference main | `df5fd2c` | https://github.com/PyAutoLabs/autolens_inference/blob/df5fd2c03cce/wiki/project/state.md | verified | +| autolens_inference row landing PR | head `fc561a9` | https://github.com/PyAutoLabs/autolens_inference/pull/23 | verified, open, not merged | +| RAL scheduler (`sacct -X`) for 350674, 350675, 350678, 350679, 350682, 350684 | 2026-10-17T11:10Z | queried by the session architect over SSH | verified | +| RAL result rows | 2026-10-17 | `/mnt/ral/jnightin/inference/autolens_inference/results/slam/imaging/hst/` (12 JSON + 12 PNG) | verified; 8 pulled by `hpc/sync` earlier, 4 `numba_cpu` copied 2026-10-17 | +| Pulled SLURM logs | 2026-10-17 | autolens_inference checkout `logs/{gpu,cpu}/output/output._.out`, each ending `Finished.` with the row path | verified (local, untracked) | +| PyAutoCortex `projects/autolens_inference.md` | origin/main `9e8b3d4` before this sweep; `ab738e4` (merged as `a61e18a`) after | https://github.com/PyAutoLabs/PyAutoCortex/blob/a61e18a/projects/autolens_inference.md | verified; six runs closed with `cortex.py done`, Now rewritten | +| PyAutoInsight | `4e9a34d` | this repository | verified | +| PyAutoMind | origin/main `11ec4317` | no `draft/*/autolens_inference` prompt and no autolens_inference claim found | verified | +| autofit_inference (gaussian-x3-search-benchmark) | not re-inspected this sweep | ledger state retained from 2026-10-16 | not re-inspected | +| graphical_ep (graphical-ep) | not re-inspected this sweep | ledger state retained from 2026-10-04 | not re-inspected | +| analytic_gaussian (analytic-gaussian) | not re-inspected this sweep | ledger state retained from 2026-10-04 | not re-inspected | +| ep_toy_gaussian (ep-toy-gaussian) | not re-inspected this sweep | ledger state retained from 2026-10-04 | not re-inspected | +| slope_hierarchy_scale (slope-hierarchy-scale) | not re-inspected this sweep | ledger state retained from 2026-10-04 | not re-inspected | +| ic50_workspace (ic50-scale) | not re-inspected this sweep | ledger state retained from 2026-10-04 | not re-inspected | + +## Campaign sweep + +| Campaign | Change since previous | Coordination state | Job evidence | Blocker | Next bounded step | +|---|---|---|---|---|---| +| slam-hst-base | All four `slam_base` re-run arrays completed; 6 rows landed in PR #23 | active | 350674, 350675 (gpu) and 350682 (ral) COMPLETED 0:0, sacct 2026-10-17T11:10Z | Three CPU legs (`jax_cpu` dense and sparse, `numba_cpu` dense) have not run; `numba_cpu` sparse walls are multi-attempt | Human reads the parity view, then decides whether the remaining CPU legs are still wanted | +| slam-hst-delaunay | Both `delaunay_1250` A100 arrays and the `numba_cpu` sparse array completed; 6 rows landed in PR #23 | active | 350678, 350679 (gpu) and 350684 (ral) COMPLETED 0:0, sacct 2026-10-17T11:10Z | Seed-0 `mass_total[1]` offset reproduces on three backends; human reading pending in Cortex | Human reads the seed-0 offset | +| hst-imaging-mass-samplers | New, opened from the user direction above | needs-decision | none submitted | Three human choices (below) and a bounded runner change | Decide the choices, then `start_dev` the runner change | +| point-source-admission | No change; project state.md 2026-10-02 entry is still the latest record | needs-decision | none new | Gradient cost share and sampler admission unmeasured | Unchanged | +| graphical-ep | Not re-inspected | needs-slicing | not queried | Unchanged | Unchanged | +| analytic-gaussian | Not re-inspected | blocked | not queried | Unchanged | Unchanged | +| ep-toy-gaussian | Not re-inspected | needs-decision | not queried | Unchanged | Unchanged | +| slope-hierarchy-scale | Not re-inspected | blocked | not queried | Unchanged | Unchanged | +| ic50-scale | Not re-inspected | blocked | not queried | Unchanged | Unchanged | +| gaussian-x3-search-benchmark | Not re-inspected | running | not queried; ledger records 409930 COMPLETED | Unchanged | Unchanged (B5c-2b) | + +Row summary for the two SLaM campaigns (search wall in seconds, chain = sum of the five stage walls; `mass_total[1]` log Z and largest absolute `truth_delta_sigma`), seed 0 / seed 1: + +| variant / config | chain | mass_total[1] wall | mass log Z | max abs Δσ | resumed stage | +|---|---|---|---|---|---| +| slam_base / A100 dense | 2,468 / 2,365 | 562 / 521 | 31,589.22 / 31,590.06 | 1.52 / 0.55 | none | +| slam_base / A100 sparse | 3,930 / 3,915 | 837 / 825 | 31,597.52 / 31,594.05 | 1.98 / 1.20 | none | +| slam_base / numba_cpu sparse | 168,824 / 177,740 | 44,221 / 50,086 | 31,592.05 / 31,595.97 | 1.30 / 3.18 | source_pix[1] | +| delaunay_1250 / A100 dense | 4,390 / 5,210 | 1,217 / 1,538 | 31,499.14 / 31,528.44 | 5.98 / 2.18 | none | +| delaunay_1250 / A100 sparse | 7,454 / 8,013 | 1,492 / 1,699 | 31,531.15 / 31,538.40 | 5.12 / 3.57 | none | +| delaunay_1250 / numba_cpu sparse | 772,484 / 757,842 | 345,641 / 360,538 | 31,515.84 / 31,544.76 | 6.62 / 3.92 | mass_total[1] | + +## Coverage and limits + +- Both `numba_cpu` arrays were requeued after their first 2026-09-24 start (the 09-24 project journal saw them RUNNING at +1 min; sacct now reports Start 2026-09-25T09:45 for 350682 and 2026-10-03T09:41 for 350684, Reason BeginTime). Each row carries one `resumed` stage, so the CPU chain walls and the resumed stage walls span more than one attempt and are not single-attempt measurements. The clean CPU stage walls are the non-resumed ones, notably `slam_base` `mass_total[1]` at 44,221 / 50,086 s. +- The A100 chain walls are single-attempt. +- Result schema v2 records no job id or library revisions. The job ids come from the pulled logs and the Cortex ledger; the library revisions (Array `3de624b5`, Lens `86054bbc1`, Fit `dd9fbe0aa`, Galaxy `70a61e26`, Nerves `1fa613a`) come from the 2026-09-24 project journal. +- One simulated target, two seeds per config. All rows report `status: complete`; that is execution completion, not convergence assessment or scientific acceptance. +- The published `inference-summary` snapshot for lens was not refreshed by this sweep; it refreshes after PR #23 merges. +- Six campaigns outside autolens_inference were not re-inspected; their ledger rows are unchanged. + +Scientific observations and any human conclusion belong in the Cortex ledger for autolens_inference; none was written this sweep. + +## Decisions and handoff + +Proposed interpretations, for the human (not rulings): + +- (a) The inference-level speedup from the likelihood work is modest on the A100: 0.98 to 1.22x on the chain across the eight legs, compared with the archived pre-speedup rows. +- (b) The `delaunay_1250` seed-0 `mass_total[1]` offset (5.98σ dense, 5.12σ sparse, 6.62σ CPU, against 2.18σ / 3.57σ / 3.92σ at seed 1) reproduces on three backends, which would answer the 2026-09-24 "confirm or retire" question as confirmed, as a property of that target and seed rather than of a backend. +- (c) `slam_base` `mass_total[1]` log Z spans about 8 nats across the six legs (31,589.2 to 31,597.5) against 1 to 4 nats seed to seed. + +Proposed for `hst-imaging-mass-samplers` (decision-gated; see `tasks/hst_mass_base_light_modes.md` and `tasks/hst_mass_samplers_wave1.md`): + +- CPU backend for the base: `jax_cpu` is recommended because the NUTS/SMC experiment path needs a JAX-frozen analysis; `numba_cpu` would instead complete the parity row. +- "Linear light removed": fixed ordinary light profiles with intensities from `light[1]` (the plan recorded in Cortex), or lens-light subtraction from the image. +- Seed count. + +Human decisions recorded after the sweep, in the Insight chat on 2026-10-17: + +- "Let's do this as a fresh start we did kind of start a campaign but weve done lots of improvements since warranting a clean slate." +- (1) Clean slate: "Yes, supersede and re-run on the pin". +- (2) CPU backend: "jax_cpu, dense + sparse". +- (3) Light mode: "Fix lens light to ordinary profiles". +- (4) Test fix: "Yes, relax the two assertions". + +Consequences recorded in the ledger: + +- `slam-hst-base` and `slam-hst-delaunay` are `superseded` by `hst-imaging-mass-samplers`. Their 2026-09-24 re-run rows stay as history in autolens_inference PR #23; no further run is planned under those campaigns. The proposed interpretations (a) to (c) above stay proposals and are not carried into the new campaign. +- `hst-imaging-mass-samplers` is reframed as a clean slate at status `needs-slicing`: the base is re-measured from scratch on PyAuto release 2026.10.17.1 (PyAutoLens tag v2026.10.17.1) in fresh output trees, on A100 `jax_gpu` dense and sparse and ral `jax_cpu` dense and sparse, seeds 0 and 1, with `mass_total[1]` run in two light modes (linear lens light, and lens light fixed to ordinary profiles with intensities from the `light[1]` fit). `numba_cpu` is not in this campaign. Phase 1 is a bounded runner change through Mind `start_dev`, not yet started; phase 2 compute needs separate authorisation. +- The task `drop_1e3_prior_centring_after_631` keeps its campaign and status and is carried into the new campaign's phase 1 as a prior-identity question for the frozen model. +- `tests/test_campaigns.py::test_current_ledger_and_migration_validate` had two migration-era assertions (`last_checkin` null, every `job_status` starting with "unknown") that this first real check-in breaks; per decision (4) they are relaxed to accept a check-in stamp with an existing report and any non-empty `job_status`. + +Authorised actions this sweep: landing the 12 rows (autolens_inference PR #23, open, not merged), closing the six runs in the Cortex ledger (`ab738e4`, merged by ledger_merge as `a61e18a`), and this report with its campaigns.yaml update. No Mind implementation issue was opened. + +Compute submitted: none. No decision above authorises compute. diff --git a/dashboard.html b/dashboard.html index 0435286..212f68c 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 2026-10-17 13:17:23 UTC
↻ Update

-

Active campaigns

CampaignRecent progressNext steps
HST SLaM base backend parityProject and Cortex ledgers retain the 24 September rerun submissions; published rows require coverage reconciliation.Reconcile all seeds and stages with current project rows and Cortex before proposing another run.
HST SLaM Delaunay variantDelaunay is a separate target/recipe; pre-speedup runs are archived, reruns recorded on 24 September.Retrieve or account for every rerun, failed seed and missing stage; preserve the archived comparison context.
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.
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.

Ledger

Inference Results

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

No decisions recorded yet.

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