diff --git a/campaigns.yaml b/campaigns.yaml index f70fe4d..722d781 100644 --- a/campaigns.yaml +++ b/campaigns.yaml @@ -162,15 +162,15 @@ tasks: - id: hst_mge_cpu_pilot campaign: setup-baseline title: Prepare and review the HST/MGE fp64 CPU pilot - status: needs-decision + status: superseded priority: high path: tasks/hst_mge_cpu_pilot.md - updated: '2026-10-13' - next: 'Review the Pulse collection choices (candidate input hashes and 31-worker resource plan verified without - execution) alongside project correctness audit #417. Resolve data origin/suitability, named host/stack/full - settings, witness/tolerances and background-load method. No freeze, compute or acceptance authorized. Post-/prm 2026-10-17: #418 merged (QUALIFICATION.md), #417 closed; record https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/hst-mge-pilot-qualification.md. Next bounded step: the synthetic witness harness that also fixes the background-load method and calibrates tolerances; still no collection, freeze or acceptance authorized.' + updated: '2026-10-25' + next: 'Superseded 2026-10-25 (human: adopt and close): the likelihood-lifecycle phase-2a imaging campaign + measured imaging/mge/hst jax_cpu_dense_fp64@hpc_ral_cpu at tag 2026.10.23.1 with parity match + (autolens_profiling#461). The pilot''s own protocol was not run. Do not reissue it.' issue: https://github.com/PyAutoLabs/autolens_profiling/issues/415 - evidence: https://github.com/PyAutoLabs/autolens_profiling/pull/416 + evidence: https://github.com/PyAutoLabs/autolens_profiling/pull/461 - id: runtime_cell_single_jit_gpu_warmup campaign: point-source-source title: Runtime cells' A100 `single_jit` includes the post-compile warm-up — source-plane 0.642 ms vs a steady diff --git a/dashboard.html b/dashboard.html index 2efad70..5b9b49e 100644 --- a/dashboard.html +++ b/dashboard.html @@ -671,8 +671,8 @@ - PyAutoPulse: https://github.com/PyAutoLabs/PyAutoPulse - autolens_profiling: https://github.com/PyAutoLabs/autolens_profiling - autofit_profiling: https://github.com/PyAutoLabs/autofit_profiling -- autolens_inference: https://github.com/PyAutoLabs/autolens_inference
Last updated 2026-10-25 16:47:53 UTC
↻ Update

-

Active campaigns

CampaignStatusLinks
Setup baseline collection and scientific acceptance
2 tasks · next step

Review the Pulse collection choices (candidate input hashes and 31-worker resource plan verified without execution) alongside project correctness audit #417. Resolve data origin/suitability, named host/stack/full settings, witness/tolerances and background-load method. No freeze, compute or acceptance authorized. Check-in 2026-10-17: no project wiki/results/issue change since 2026-10-13 beyond autolens_profiling#420/#422/#424 (interferometer only); releases 2026.10.16.1/.2 and 2026.10.17.1 supply no new measurement. Post-/prm 2026-10-17: autolens_profiling#418 (pilot qualification design) merged after an independent review (FINDINGS, none blocking); issue #417 closed; Mind record complete/2026/10/hst-mge-pilot-qualification.md. Carried to the next bounded step: background-load method (none proposed), the 1e-8 relative image L2 budget (likely unattainable; synthetic-harness calibration must be allowed to revise tolerances), and the correctness-process feasibility needed before the 32-process budget is final.

Reviewed 2026-10-17

Active tasks

Freeze the setup baseline specification, then collect and review evidence↗

needs-decision · high

HST/MGE raw fixed-parameter CPU runner merged in #416; Pulse handoff #55 merged and Mind implementation closed. Resolve exact production host/stack/data, full settings, independent correctness witness/tolerances and background-load method. No collection, campaign freeze or acceptance authorized; global compile dispatch remains unavailable.

Prepare and review the HST/MGE fp64 CPU pilot↗

needs-decision · high

Review the Pulse collection choices (candidate input hashes and 31-worker resource plan verified without execution) alongside project correctness audit #417. Resolve data origin/suitability, named host/stack/full settings, witness/tolerances and background-load method. No freeze, compute or acceptance authorized. Post-/prm 2026-10-17: #418 merged (QUALIFICATION.md), #417 closed; record https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/hst-mge-pilot-qualification.md. Next bounded step: the synthetic witness harness that also fixes the background-load method and calibrates tolerances; still no collection, freeze or acceptance authorized.

needs-decision
↗☷
Point-source image plane · CPU
1 task · next step

No new committed campaign measurements. Phases 1–4 remain shipped; epic closure is a human decision. Carried leftovers remain pending. RAL queue empty at review, but transfer preview timed out: cleanup folders and unpulled results unverified. Select one owed cleanup/measurement phase only after review. Check-in 2026-10-17: no project wiki/results/issue change since 2026-10-13 beyond autolens_profiling#420/#422/#424 (interferometer only); releases 2026.10.16.1/.2 and 2026.10.17.1 supply no new measurement. RAL queue holds only euclid_dr1 jobs; cleanup folders and unpulled results still unverified.

Reviewed 2026-10-17

Active tasks

active
↗☷
Point-source image plane · A100
1 task · next step

Parked as research, human decision 2026-10-17 (chat) ("just research to file"). Reference measurement autolens_profiling#434 (issue #427 closed) merged 2026-10-17 (human /prm): job 412519 euclid-ral-gpu-2 exclusive A100, jax 0.10.2, fp64, 2026.10.17.1 private clone: single call 0.934 ms (pre-#768 0.910), vmap-256 0.0125 ms (0.0120), compile 6.54 s; forward gradient finite 16/16, equal to reverse to 1.6e-11; logL bit-exact; +2.7% not attributed (different node/date). A100 per-step breakdown rows unusable. Phase-2 inference fit measurement dropped; no baseline admitted. Records: https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/point-source-image-gpu-admission.md.

Reviewed 2026-10-17

Active tasks

Point-source A100 speed-up campaign: profile and optimize with the shared breakdown↗

parked · high

Parked, human decision 2026-10-17 (chat). Reference A100 measurement autolens_profiling#434 (merged 2026-10-17 (human /prm), issue #427 closed, job 412519): 0.934 ms single call, 0.0125 ms at vmap-256, forward gradient fixed; phase-2 inference fit measurement dropped. Records: https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/point-source-image-gpu-admission.md.

parked
↗☷
Cluster PointSolver robustness and performance
2 tasks · next step

Human decision 2026-10-23 (chat): keep the 1.0″ starting grid as the production default (decision record cluster-pointsolver-starting-grid; closes the phase-1 deferral). Inputs: phase-1 simple sweep (#438), A2744 pilot job 412573 (1.0″ misses fewest, 2/69), starting-grid follow-up autolens_profiling#439 → #440 merged 03d36c81 (no fixed scale finds every image; no separation rule). Mitigation: the workspace validation check. Next bounded task: spurious-core candidates library task (start_dev in progress), then the validation check. No default changed.

Reviewed 2026-10-23

Active tasks

active
↗☷
Interferometer likelihood and streaming
5 tasks · next step

Evening 2026-10-17: human decision 2026-10-17 (chat): interferometer CPU work is exhaustive and closed; curvature preload parked ("we will come back to it soon"); GPU levers "file wiki research but not make source code changes for now", so no library prompt for JAX FFT pruning. Parked at knowledge-complete. All merged 2026-10-17 (human /prm), pending release for the library PRs: L2 symmetric direct_conv PyAutoArray#636 (issue #635 closed, pending-release; real Aris RAL 1.49×/1.53×/1.58× at 1,459/2,577/5,746 vertices, logL to round-off; decision record direct-conv-symmetric-gate) + record autolens_profiling#425; A100 FFT levers #433 (issue #429 closed) (pruned FFT 1.08×/1.18×/1.41× whole likelihood at alma/alma_high/jvla; pixel-scale coarsening fails the 0.5-nat bar); streaming 1e8 CPU/A100 + memo after-measurement #435 (issues #368, #426 closed). Route-constant refit (RAL FFT/full direct_conv 0.68 vs 0.89 predicted) deferred. Merged in order #636, #425, #432, #431, #433, #434, #435. Records: https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/direct-conv-symmetric.md, https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/ral-route-constants-fft-levers.md, https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/interferometer-streaming-memo-after.md.

Reviewed 2026-10-17

Active tasks

fnnls warm-start memo: stop it slowing scattered evaluation streams (interferometer CPU, autolens_profiling#332)↗

parked · medium

After-measurement autolens_profiling#435 merged 2026-10-17 (human /prm) (issue #426 closed; job 412526, ral compute-10-2, load ~10, 1 thread, 6 rounds, paired bootstrap 90% CI): Delaunay iid #332 flags 1.918 [1.887,1.925] (2.17 in #332); 15 timed solves 1.018 [1.013,1.023]; Delaunay walk 0.46; rect iid 0.887/0.987; rect walk 0.15; FoM identical. Not evidence #613 is fixed. Research owed: Nautilus-replay witness, varied-seed rounds. Parked, human decision 2026-10-17 (chat). Records: https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/interferometer-streaming-memo-after.md.

Campaign: interferometer streaming (array-free) vs in-memory — memory and time scaling to 2e8 visibilities↗

parked · high

Phase 2 autolens_profiling#435 merged 2026-10-17 (human /prm) (issue #368 closed): from_stream chunk 65536 fp64, 1e8 vis RAL CPU 707.5 s / 1.78 GB RSS, A100 157.3 s / 1.48 GB device; compile ~3 s constant. In-memory fails at 2.5e5 (RAL 10 GB cap) and 4e6 (A100); uncapped CPU ~7.5 kB/vis. Optional 2e8 rows are research; library candidate (chunk transformer.image_from) unfiled per human decision 2026-10-17 (chat). Parked. Records: https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/interferometer-streaming-memo-after.md.

Interferometer large-mask likelihood (Aris 512×512): sparse data-vector memory, AdaptSplit log-det backend parity, numba route-gate cost model↗

active · high

L2 shipped to PR: PyAutoArray#636 (head 62a1e3ac, pending-release) + autolens_profiling#425 (results note), both merged 2026-10-17 (human /prm); issue #635 closed. RAL real-data witness 1.49×/1.53×/1.58× at 1,459/2,577/5,746 vertices (pred 1.48×/1.50×/1.53×), logL to round-off; peak RSS 43 GB without NUFFT chunking. Decision record direct-conv-symmetric-gate added. Route-constant refit (0.68 measured vs 0.89 predicted) deferred: CPU lane closed, human decision 2026-10-17 (chat). Remaining: release of #636; shadow row owed in Mind autonomy_log (glance tier; human to confirm merged-unchanged). Records: https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/direct-conv-symmetric.md. Shadow row appended 2026-10-17 (merged-unchanged, stage 1). Human decision 2026-10-17: remaining Aris levers stay research; not enough return to implement for now.

Interferometer W~ curvature matrix is FFT-bound on the mask extent: pruned padded FFT and real-space pixel scale on the A100

parked · normal

A100 research autolens_profiling#433 merged 2026-10-17 (human /prm) (issue #429 closed; job 412515): pruned padded FFT 1.08×/1.18×/1.41× whole likelihood at alma/alma_high/jvla, F exact to 1.7e-15; pixel-scale coarsening shifts log evidence by +1,310 to +2,757 nats (fails). No library prompt, human decision 2026-10-17 (chat). Parked. Records: https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/ral-route-constants-fft-levers.md.

parked
↗☷
Linear-solver accuracy and cost
0 tasks · next step

Retain the 2026-10-08 human decision: Jacobi for Mapper, raw+polish for MGE-only, no default change. Project wiki still says decision pending (stale). No new solver measurements after the recorded phase-5 results; release 2026.10.12.1 does not supply a rerun. Wide-prior/interferometer gaps remain evidence gaps, not authorized tasks. Check-in 2026-10-17: no project wiki/results/issue change since 2026-10-13 beyond autolens_profiling#420/#422/#424 (interferometer only); releases 2026.10.16.1/.2 and 2026.10.17.1 supply no new measurement.

Reviewed 2026-10-17

Active tasks

No open tasks.

active
↗☷
Production imaging and likelihood breakdown
3 tasks · next step

Production over-sampling phase 1 merged 2026-10-17 (human /prm): autolens_profiling#431 (issue #430 closed, CPU smoke 8/8): --over-sampling {flat1,production} on the 4 JAX Delaunay/DelaunayNN cells, euclid/hst presets. Findings: HST flat-1 adapt_split pin fails on unmodified main (29155.000 vs 29140.296); numba cells' production threshold never reaches sub 4 for HST; rectangular JAX cells uncovered. Phase 2 (A100 production re-measure) needs-decision (human decision 2026-10-17 (chat): undecided). DelaunayNN imaging moved to the parked delaunay-nn-cpu campaign. Post-certified, quick-update plotting and other tasks unchanged. Records: https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/imaging-production-over-sampling.md. Human decision 2026-10-17 (chat): hold off on phase 2 (A100 production over-sampling re-measure); not authorized, revisit when production imaging needs a cost picture.

Reviewed 2026-10-17

Active tasks

Profile imaging pixelizations at production over-sampling

parked · normal

Phase 1 merged 2026-10-17 (human /prm): autolens_profiling#431 (issue #430 closed, CPU smoke 8/8). Phase 2 A100 production re-measure needs-decision (human decision 2026-10-17 (chat): undecided); requires compute authorization (#431 now merged). Findings: HST flat-1 adapt_split pin fails on main; numba HST threshold never reaches sub 4; rectangular JAX cells uncovered. Records: https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/imaging-production-over-sampling.md. Human decision 2026-10-17 (chat): phase 2 A100 re-measure held; task stays parked at phase 1 shipped (#431).

active
↗☷
DelaunayNN (Sibson) mesh at Delaunay cost on the numba/NumPy CPU path
1 task · next step

Created parked, human decision 2026-10-17 (chat): "file in research to come back to in the future". PyAutoArray#637 (issue #629 closed) merged 2026-10-17 (human /prm), pending-release: RAL compute-10-4 exclusive 1 thread A-B-B-A ×2, 1,030 vertices / 15,380 px: NN sub4 0.569→0.392 s (2.20× Delaunay, was 3.00×), sub2 0.235→0.189 s (1.75×); logL bit-identical; fixes the size-0 Sibson NaN case. ~1.5× target not met; remaining cost is Sibson (~0.16 s) + wider curvature rows. Target and two hazards recorded as future work, not filed. Records: https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/delaunaynn-wide-row-overhead.md.

Reviewed 2026-10-17

Active tasks

DelaunayNN (Sibson) imaging likelihood at close to Delaunay cost on the numba/NumPy CPU path↗

parked · normal

Parked, human decision 2026-10-17 (chat). PyAutoArray#637 merged 2026-10-17 (human /prm) (issue #629 closed, pending-release): sub4 2.20× Delaunay (was 3.00×), sub2 1.75×; logL bit-identical. Future (unfiled): ~1.5× target; hazards sub_slim_indexes_for_pix_index padding and Sibson rows with mid-row gaps. Records: https://github.com/PyAutoLabs/PyAutoMind/blob/main/complete/2026/10/delaunaynn-wide-row-overhead.md.

parked
↗☷
Measurement reliability and profiling tools
8 tasks · next step

Timing audit #410–#412 and raw HST/MGE pilot producer #416 merged; implementation issues #362/#415 closed. No fresh production measurements or scientific acceptance. C1/C11 repeat counts, C5 estimator policy, memo interpretation, fresh runtime repeats and other tools remain pending. #418 and #417 open. Evening 2026-10-17 merge-order note: Six open autolens_profiling PRs (#425, #431, #432, #433, #434, #435) all regenerate the shared catalogue/dashboard: each merge after the first needs a rebase + re-render; suggested order #425, #432, #431, #433, #434, #435. #425 after PyAutoArray#636.

Reviewed 2026-10-17

Active tasks

jax_compile/probe.py lost its cell builder with the searches tier — give profiling its own

active · high

Manual HST/MGE fp64 jit builder #414 and separate raw pilot runner #416 merged. Bounded Mind implementations complete; wider builder coverage/global compile dispatch still unavailable. Agree a next capability or separately authorize a qualified pilot; no production compute authorized. Check-in 2026-10-17: autolens_profiling#418 (pilot qualification design) open since 2026-10-13 with CI green and no review; #417 open; no new measurements.

active
↗☷
PyAutoFit profiling bootstrap
1 task · next step

B4a (autofit_profiling#3) and B4b (autofit_profiling#5, issue #4 closed, merged 2026-10-15) shipped: search.fit breakdown cell, profiling-summary@2 exporter, EP/graphical N3/N10/N30 baselines and analytic parity cell; #6 repointed RAL paths to the category layout (2026-10-15). Pulse fit fixture/campaign merged (#60, #61); fit receipt ok at autofit_profiling 58e56b9 (31 records, fetched 2026-10-15T22:58Z), so the post-deploy refresh is done. Epic 2 not opened: next bounded step is a human-scoped epic-2 campaign row; B5c inference work belongs to Insight, not Pulse.

Reviewed 2026-10-17

Active tasks

autofit_profiling: bootstrap the repo + general PyAutoFit profiling epic

active · normal

B4a (autofit_profiling#3) and B4b (autofit_profiling#5, issue #4 closed, merged 2026-10-15) shipped: search.fit breakdown cell, profiling-summary@2 exporter, EP/graphical N3/N10/N30 baselines and analytic parity cell; #6 repointed RAL paths to the category layout (2026-10-15). Pulse fit fixture/campaign merged (#60, #61); fit receipt ok at autofit_profiling 58e56b9 (31 records, fetched 2026-10-15T22:58Z), so the post-deploy refresh is done. Epic 2 not opened: next bounded step is a human-scoped epic-2 campaign row; B5c inference work belongs to Insight, not Pulse.

active
↗☷
Critical curves and evaluation grids
0 tasks · next step

No new campaign measurement. Phase 3b cap fix remains shipped; project wiki still incorrectly says #646 awaits merge/release. Seed coverage and path completion remain unresolved; retain marching-squares decision and ask the human whether to scope the next phase. Check-in 2026-10-17: no project wiki/results/issue change since 2026-10-13 beyond autolens_profiling#420/#422/#424 (interferometer only); releases 2026.10.16.1/.2 and 2026.10.17.1 supply no new measurement.

Reviewed 2026-10-17

Active tasks

No open tasks.

needs-decision
↗☷
Certified positive solver
0 tasks · next step

Parked unchanged. No new campaign evidence; existing C2/library/workspace drafts remain pending. Do not infer admission from solver timing, and do not change default or restart the campaign. Check-in 2026-10-17: no project wiki/results/issue change since 2026-10-13 beyond autolens_profiling#420/#422/#424 (interferometer only); releases 2026.10.16.1/.2 and 2026.10.17.1 supply no new measurement.

Reviewed 2026-10-17

Active tasks

No open tasks.

parked
↗☷

Open means tracked, not necessarily running. Full ledger ↗

Profiling Results

PyAutoLens: setups by class

300 current · 379 superseded (shown with History) · 99 archived (counted, not shown). Class and status are the producer's declarations; this board ranks nothing.

LikelihoodInstrumentRegimeSetupHardware classModel variantMeasurementValue (producer)ClassStatusDecisionEvidence

No record matches this selection.

Coverage: 47 optimal cells not measured
  • cluster/image_plane · simple · jax_gpu_na_fp64 · hpc_a100 · default · runtime
  • cluster/image_plane · simple · jax_gpu_na_fp64 · hpc_a100 · default · breakdown
  • cluster/image_plane · simple · jax_cpu_na_fp64 · hpc_ral_cpu · default · runtime
  • cluster/image_plane · simple · jax_cpu_na_fp64 · hpc_ral_cpu · default · breakdown
  • datacube/delaunay · alma · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · runtime
  • datacube/delaunay · alma · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · breakdown
  • datacube/delaunay · alma · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · runtime
  • datacube/delaunay · alma · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · breakdown
  • datacube/delaunay · alma_high · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · runtime
  • datacube/delaunay · alma_high · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · breakdown
  • datacube/delaunay · alma_high · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · runtime
  • datacube/delaunay · alma_high · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · breakdown
  • datacube/delaunay · sma · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · runtime
  • datacube/delaunay · sma · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · breakdown
  • datacube/delaunay · sma · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · runtime
  • datacube/delaunay · sma · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · breakdown
  • imaging/delaunay · ao · jax_gpu_dense_fp64 · hpc_a100 · default · breakdown
  • imaging/rectangular · ao · jax_gpu_dense_fp64 · hpc_a100 · default · breakdown
  • interferometer/delaunay · alma · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · runtime
  • interferometer/delaunay · alma · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · runtime
  • interferometer/delaunay · alma_high · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · runtime
  • interferometer/delaunay · alma_high · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · runtime
  • interferometer/delaunay · jvla · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · runtime
  • interferometer/delaunay · jvla · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · runtime
  • interferometer/delaunay · jvla · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · breakdown
  • interferometer/delaunay · sma · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · runtime
  • interferometer/delaunay · sma · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · runtime
  • interferometer/mge · sma · jax_gpu_dense_fp64_nufft · hpc_a100 · default · runtime
  • interferometer/mge · sma · jax_cpu_dense_fp64_nufft · hpc_ral_cpu · default · runtime
  • interferometer/mge · sma · jax_cpu_dense_fp64_nufft · hpc_ral_cpu · default · breakdown
  • interferometer/rectangular · alma · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · runtime
  • interferometer/rectangular · alma · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · runtime
  • interferometer/rectangular · alma_high · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · runtime
  • interferometer/rectangular · alma_high · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · runtime
  • interferometer/rectangular · jvla · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · runtime
  • interferometer/rectangular · jvla · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · runtime
  • interferometer/rectangular · jvla · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · breakdown
  • interferometer/rectangular · sma · jax_gpu_sparse_fp64_wtilde · hpc_a100 · default · runtime
  • interferometer/rectangular · sma · numba_cpu_sparse_fp64_wtilde · hpc_ral_cpu · default · runtime
  • multi_dataset/delaunay · hst · jax_gpu_dense_fp64 · hpc_a100 · default · runtime
  • multi_dataset/delaunay · hst · jax_gpu_dense_fp64 · hpc_a100 · default · breakdown
  • multi_dataset/delaunay · hst · numba_cpu_sparse_fp64 · hpc_ral_cpu · default · runtime
  • multi_dataset/delaunay · hst · numba_cpu_sparse_fp64 · hpc_ral_cpu · default · breakdown
  • point_source_image/image_plane · simple · jax_gpu_na_fp64 · hpc_a100 · default · runtime
  • point_source_image/image_plane · simple · jax_cpu_na_fp64 · hpc_ral_cpu · default · runtime
  • point_source_source/source_plane · simple · jax_cpu_na_fp64 · hpc_a100 · default · runtime
  • point_source_source/source_plane · simple · jax_cpu_na_fp64 · hpc_a100 · default · breakdown

autolens_inference: wall-time validation

Full-fit measurements: one sampler run per row. Every derived number is the producer's; Pulse computes no ratio and never reads a full-fit time as a per-likelihood measurement.

LikelihoodInstrumentSetupHardware classModel variantSamplerSeedCoresWall sLikelihood evalsImplied ms/evalPredicted ms/evalMeasured/predictedParallel efficiencyParityReference runtime recordRelease
imaging/delaunayhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:mass_total[1]188648.19728112220838660223.69884327786363———not_assessed—2026.10.17.1
imaging/delaunayhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:source_pix[2]18453.90562796592712120214.1064282858147———not_assessed—2026.10.17.1
imaging/delaunayhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:source_pix[1]1812609.19023084640553920233.84996718928795———not_assessed—2026.10.17.1
imaging/delaunayhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:light[1]182269.4144601821910180222.928728898054———not_assessed—2026.10.17.1
imaging/delaunayhstnumba_cpu_sparse_fp64hpc_ral_cpulight_fixednautilus:mass_total[1]184563.57740211486834060133.98641814782349———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:mass_total[1]0832440.87102484703304001067.1339152910207———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:source_pix[2]085783.56625437736555601040.2097579815406———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:source_pix[1]0838212.712732076645345401106.3321578481946———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:light[1]085446.458575010311080491.5576331236732———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:source_lp[1]083367.14607405662549255036.38191327992032———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:mass_total[1]1824129.204991340637220201095.7858760826812———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:source_pix[2]184042.28769350051937201086.6364767474513———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:source_pix[1]1838641.76520562172341801130.5373085319404———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:light[1]185703.01235604286210620537.0068131867101———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:source_lp[1]183410.5005660057079125037.375348668555695———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_sparse_fp64hpc_a100defaultnautilus:mass_total[1]0836050.4700050354217201659.78222859279———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_sparse_fp64hpc_a100defaultnautilus:source_pix[2]086332.76647329330438201657.7922704956293———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_sparse_fp64hpc_a100defaultnautilus:source_pix[1]0883132.6692700386345402406.852034453926———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_sparse_fp64hpc_a100defaultnautilus:light[1]083972.18898868560810880365.0908996953684———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_sparse_fp64hpc_a100defaultnautilus:source_lp[1]083471.95770215988169255037.51439980723805———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:mass_total[1]04558.61791110038762368023.590283407955553———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:source_pix[2]04181.06282258033752780023.21318238209455———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:source_pix[1]04700.86670279502873302021.225520981072947———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:light[1]04234.46286773681641160020.21231618420831———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:source_lp[1]04714.9393587112427925507.724898527404027———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:mass_total[1]14524.9951443672182164024.26040408351285———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:source_pix[2]14130.3421642780304500026.06843285560608———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:source_pix[1]14732.56710720062263452021.22152685980946———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:light[1]14219.777843713760381010021.76018254591687———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:source_lp[1]14712.3624505996704877008.122718935001943———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:mass_total[1]04826.86189579963682306035.85697726798078———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:source_pix[2]04298.14068365097046814036.62661961314134———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:source_pix[1]041786.15557265281683282054.42277796017114———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:light[1]04221.93183565139771110019.993859067693485———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:source_lp[1]04714.965235710144925507.725178127608256———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:mass_total[1]14811.51361227035522298035.31390827982399———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:source_pix[2]14226.34285354614258600037.723808924357094———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:source_pix[1]141648.6028597354893036054.30180697415971———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:light[1]14196.74390411376953920021.385206968887992———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:source_lp[1]14719.88591837883877008.208505340693614———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:mass_total[1]082605.16190314292922920113.66325929942973———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:source_pix[2]08579.78667187690735660102.43580775210377———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:source_pix[1]084961.52095890045233660147.40109800654935———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:light[1]081275.3287878036510520121.22897222468156———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:mass_total[1]182562.109365701675422240115.20275924917605———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:source_pix[2]18939.1733293533325946099.278364625088———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:source_pix[1]184867.77173519134533200146.61963057805258———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:light[1]181478.910149812698412000123.24251248439153———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100light_fixednautilus:mass_total[1]0821738.9960973262823780914.1714086344103———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100light_fixednautilus:mass_total[1]1820092.07260394096420940959.506810121345———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100light_fixednautilus:mass_total[1]04490.452539205551152282021.492223453354562———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100light_fixednautilus:mass_total[1]14449.707456111907962078021.64135977439403———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100light_fixednautilus:mass_total[1]041624.85360050201424870033.36455031831651———diverged—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100light_fixednautilus:mass_total[1]141396.9158349037174188033.35520140648799———diverged—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpulight_fixednautilus:mass_total[1]082329.36153244972232434095.70096682209213———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpulight_fixednautilus:mass_total[1]182076.246310234072196094.54673543870992———not_assessed—2026.10.17.1

Decision History

Keep the 1.0″ PointSolver starting grid for cluster fits, with a validation checkGate the symmetric direct_conv kernel at work fraction 0.7, with ymin column orderingRoute interferometer curvature by a calibrated two-constant cost modelExclude zero-weight mappings from AdaptSplit pixel signals on Delaunay meshes onlyKeep Jacobi PDIP as the Mapper default and raw+polish as the MGE-only default

autolens_inference: wall-time validation

Full-fit measurements: one sampler run per row. Every derived number is the producer's; Pulse computes no ratio and never reads a full-fit time as a per-likelihood measurement.

LikelihoodInstrumentSetupHardware classModel variantSamplerSeedCoresWall sLikelihood evalsImplied ms/evalPredicted ms/evalMeasured/predictedParallel efficiencyParityReference runtime recordRelease
imaging/delaunayhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:mass_total[1]188648.19728112220838660223.69884327786363———not_assessed—2026.10.17.1
imaging/delaunayhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:source_pix[2]18453.90562796592712120214.1064282858147———not_assessed—2026.10.17.1
imaging/delaunayhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:source_pix[1]1812609.19023084640553920233.84996718928795———not_assessed—2026.10.17.1
imaging/delaunayhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:light[1]182269.4144601821910180222.928728898054———not_assessed—2026.10.17.1
imaging/delaunayhstnumba_cpu_sparse_fp64hpc_ral_cpulight_fixednautilus:mass_total[1]184563.57740211486834060133.98641814782349———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:mass_total[1]0832440.87102484703304001067.1339152910207———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:source_pix[2]085783.56625437736555601040.2097579815406———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:source_pix[1]0838212.712732076645345401106.3321578481946———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:light[1]085446.458575010311080491.5576331236732———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:source_lp[1]083367.14607405662549255036.38191327992032———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:mass_total[1]1824129.204991340637220201095.7858760826812———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:source_pix[2]184042.28769350051937201086.6364767474513———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:source_pix[1]1838641.76520562172341801130.5373085319404———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:light[1]185703.01235604286210620537.0068131867101———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100defaultnautilus:source_lp[1]183410.5005660057079125037.375348668555695———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_sparse_fp64hpc_a100defaultnautilus:mass_total[1]0836050.4700050354217201659.78222859279———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_sparse_fp64hpc_a100defaultnautilus:source_pix[2]086332.76647329330438201657.7922704956293———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_sparse_fp64hpc_a100defaultnautilus:source_pix[1]0883132.6692700386345402406.852034453926———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_sparse_fp64hpc_a100defaultnautilus:light[1]083972.18898868560810880365.0908996953684———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_sparse_fp64hpc_a100defaultnautilus:source_lp[1]083471.95770215988169255037.51439980723805———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:mass_total[1]04558.61791110038762368023.590283407955553———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:source_pix[2]04181.06282258033752780023.21318238209455———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:source_pix[1]04700.86670279502873302021.225520981072947———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:light[1]04234.46286773681641160020.21231618420831———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:source_lp[1]04714.9393587112427925507.724898527404027———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:mass_total[1]14524.9951443672182164024.26040408351285———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:source_pix[2]14130.3421642780304500026.06843285560608———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:source_pix[1]14732.56710720062263452021.22152685980946———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:light[1]14219.777843713760381010021.76018254591687———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100defaultnautilus:source_lp[1]14712.3624505996704877008.122718935001943———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:mass_total[1]04826.86189579963682306035.85697726798078———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:source_pix[2]04298.14068365097046814036.62661961314134———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:source_pix[1]041786.15557265281683282054.42277796017114———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:light[1]04221.93183565139771110019.993859067693485———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:source_lp[1]04714.965235710144925507.725178127608256———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:mass_total[1]14811.51361227035522298035.31390827982399———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:source_pix[2]14226.34285354614258600037.723808924357094———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:source_pix[1]141648.6028597354893036054.30180697415971———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:light[1]14196.74390411376953920021.385206968887992———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100defaultnautilus:source_lp[1]14719.88591837883877008.208505340693614———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:mass_total[1]082605.16190314292922920113.66325929942973———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:source_pix[2]08579.78667187690735660102.43580775210377———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:source_pix[1]084961.52095890045233660147.40109800654935———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:light[1]081275.3287878036510520121.22897222468156———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:mass_total[1]182562.109365701675422240115.20275924917605———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:source_pix[2]18939.1733293533325946099.278364625088———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:source_pix[1]184867.77173519134533200146.61963057805258———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpudefaultnautilus:light[1]181478.910149812698412000123.24251248439153———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100light_fixednautilus:mass_total[1]0821738.9960973262823780914.1714086344103———not_assessed—2026.10.17.1
imaging/rectangularhstjax_cpu_dense_fp64hpc_a100light_fixednautilus:mass_total[1]1820092.07260394096420940959.506810121345———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100light_fixednautilus:mass_total[1]04490.452539205551152282021.492223453354562———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_dense_fp64hpc_a100light_fixednautilus:mass_total[1]14449.707456111907962078021.64135977439403———not_assessed—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100light_fixednautilus:mass_total[1]041624.85360050201424870033.36455031831651———diverged—2026.10.17.1
imaging/rectangularhstjax_gpu_sparse_fp64hpc_a100light_fixednautilus:mass_total[1]141396.9158349037174188033.35520140648799———diverged—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpulight_fixednautilus:mass_total[1]082329.36153244972232434095.70096682209213———not_assessed—2026.10.17.1
imaging/rectangularhstnumba_cpu_sparse_fp64hpc_ral_cpulight_fixednautilus:mass_total[1]182076.246310234072196094.54673543870992———not_assessed—2026.10.17.1

Decision History

Keep the 1.0″ PointSolver starting grid for cluster fits, with a validation checkGate the symmetric direct_conv kernel at work fraction 0.7, with ymin column orderingRoute interferometer curvature by a calibrated two-constant cost modelExclude zero-weight mappings from AdaptSplit pixel signals on Delaunay meshes onlyKeep Jacobi PDIP as the Mapper default and raw+polish as the MGE-only default