Measure. Trace. Compare.
| Campaign | Status | Links |
|---|---|---|
Point-source image plane · CPU1 task · next stepPhases 1–4 shipped; PyAutoLens#764 (extent warning) released in 2026.10.4.1. Next: docs-only autolens_profiling PR writing campaign completion evidence and fixing stale wiki/index lines (#763 'in flight' → #764 merged). RAL leftover folders present; mirror sync unverified. Active tasksPoint-source (single-source) CPU campaign — carried leftovers and completion evidence↗Execution 2026-10-04: docs-only branch feature/point-source-wiki-reconcile @0560284 (issue autolens_profiling#370), parked at Heart RED. Still owed: RAL cleanup, register_model grad-zero prompt, test move, CI smoke cells, nopad deletion. | active | |
Point-source image plane · A1001 task · next stepPhase 0+1 shipped (profiling#353 merged 2026-09-30; wiki stale). Forward-mode gradient NaN is unfixed and live in 2026.10.4.1 but only a Mind draft. Human decisions: issue that bug; authorize the autolens_inference image-plane fit measurement. Active tasksPoint-source A100 speed-up campaign: profile and optimize with the shared breakdownExecution 2026-10-04: forward-mode NaN issued as PyAutoLens#767 (reproduced on CPU fp64: forward [nan]x5, reverse finite; stack 2026.10.4.1+3 b695e57b6); Mind registry 0112ccba. The autolens_inference image-plane fit measurement still awaits human authorisation. | needs-decision | |
Point-source source plane2 tasks · next stepCore complete. Nautilus leaf autolens_inference#17 merged 2026-10-02 (wiki still says open). Blackjax parked: gradient-sampler admission not met. Warm-up task awaits the human's option (a)/(b)/(c). Active tasksRuntime cells' A100 `single_jit` includes the post-compile warm-up — source-plane 0.642 ms vs a steady 0.267 ms on the same node↗Execution 2026-10-04: human chose option (a). Branch feature/runtime-single-jit-median @87a7fcc (issue autolens_profiling#371) parked at Heart RED. Local CPU witness: single_jit 0.417 ms, median 0.388 ms (p10 0.320, p90 0.545). Imaging cells not yet wired. Point-source source-plane chi-squared speed-up campaign — remaining candidates (phases 1–2e shipped)Still blocked: gradient-sampler admission not met (autolens_inference has only the Nautilus source-plane leaf, #17 merged 2026-10-02). Blackjax stays parked. | parked | |
Cluster PointSolver robustness and performance2 tasks · next stepUnchanged since 2026-09-27; nothing issued or claimed. Wiki still points at the removed Mind draft (now Pulse tasks/). Pick one bounded phase: observable-overflow invalid-result policy + CI witness, or a representative-data baseline cell. Active tasksCluster PointSolver — robustness, analysis settings and performanceUnissued, no claim. Pick one bounded phase per Mind epics.md: observable-overflow invalid-result policy + CI witness, or a representative-data baseline cell. PointSolver profiling cells: lensed quasar → cluster runtime tier → single/multi-source → multiplaneUnissued, no claim. Candidate first cell: a representative-data cluster likelihood_runtime baseline (single-source preset). | active | |
Interferometer likelihood and streaming5 tasks · next stepLast CPU cell ran (RAL job 375978_3 COMPLETED 1:11:36, JSON on RAL, not on main) but misses the ≤1e-3 nat CPU/A100 witness: human decision needed. Streaming phases 3–5 shipped; the scaling measurement needs compute authorization. Curvature, memo and FFT tasks unstarted. Active tasksInterferometer decision matrix — fill the last cell (CPU rect 39² at alma_high r5.0) and flip the wiki row to shipped↗Execution 2026-10-04: human decision ACCEPT the 1.59e-3 nat CPU/A100 gap (precedent r3.5 at 1.5e-3). Branch feature/interferometer-decision-matrix-last-cell @93a002a committed locally; ship PARKED at Heart RED (release validation FAILED, stage integrate). Caveat: the A100 row's own A/B shows certified solver = PDIP to 0.0 nat, so the gap may be F/D on older revisions rather than solver. Correction: PyAutoArray#582 first released 2026.9.27.2. Next: ship once Heart clears. Interferometer fixed-mapper searches: reuse the W~ curvature matrix across likelihood calls via `preloads.curvature_matrix`Phase 1 sliced 2026-10-04 into a Mind draft (PyAutoArray preload plumbing + CPU breakdown lever, no GPU); run it through start_dev. Human science call first - no production SLaM stage holds the mapper fixed today. fnnls warm-start memo: stop it slowing scattered evaluation streams (interferometer CPU, autolens_profiling#332)Unstarted (no PR/issue/branch/results, 2026-10-04). Select one bounded step when prioritised. Campaign: interferometer streaming (array-free) vs in-memory — memory and time scaling to 2e8 visibilities↗Execution 2026-10-04: branch feature/interferometer-streaming-scaling @4a0ef46 (issue autolens_profiling#368) parked at Heart RED. Laptop CPU 8-thread indicative rows recorded (chunk 65536 to 5e7 vis; in-memory first failure 1e6 under a 10 GB cap; parity 1.2e-9 nats at 5e5; 1e8 skipped). Go/no-go overtaken; recorded as release evidence for 2026.10.4.1. nufft_chunk_size arm hits a second memory wall in transformer.image_from (library candidate, no issue filed). Interferometer W~ curvature matrix is FFT-bound on the mask extent: pruned padded FFT and real-space pixel scale on the A100Unstarted (no PR/issue/branch/results, 2026-10-04). Select one bounded step when prioritised. | active | |
Linear-solver accuracy and cost1 task · next stepRelease half met: PyAutoArray#595 is in 2026.10.2.1 and 2026.10.4.1. RAL mirror sync (HPCPullPyAuto, plus a dependency-floor diff) unverified; phase 3 GPU/vmap parity and timing waits on it. Active tasksLinear-solver programme phase 3: GPU/vmap/A100 timing and parity rows for the solver corpusRAL sync SKIPPED 2026-10-04: 180 euclid_dr1 tasks RUNNING on the shared base; mirror at 2026-09-27 state, 16–35 commits behind tag 2026.10.4.1; a sync pulls MAINS and also needs a jax reinstall (Nerves main excludes jax 0.10.2). Next: sync when `squeue -u jnightin -t R` shows no euclid_dr1 jobs on the shared base, or the human accepts the risk. | blocked | |
Production imaging and likelihood breakdown3 tasks · next stepNo new results since the pin. Over-sampling phase 1: add --instrument production presets to the JAX Delaunay cells (CPU smoke only); PyAutoArray#606 S/N over-sampling helper (in 2026.10.4.1) is new context. Post-certified Blocked-by is stale (resolved 2026-09-27). Active tasksProfile imaging pixelizations at production over-samplingPhase 1 (workspace-only): --instrument euclid|hst presets for the JAX Delaunay runtime/breakdown cells, keeping flat-1 selectable; CPU smoke only. A100 re-measure is a separate compute-authorized phase. Assess remaining likelihood bottlenecks after certified solver integration on CPU and GPUsBlocked-by is stale: resolved 2026-09-27 (PyAutoArray#567 in 2026.9.26.1). Unissued; scope one bounded phase from the full task contract. Quick-update plotting cost — minutes per update, and it is not JAX compileUnchanged; targets PyAutoLens/PyAutoGalaxy (PyAutoLens#680 closed 2026-07-31, no follow-up issue). Select one bounded step when prioritised. | active | |
Measurement reliability and profiling tools9 tasks · next stepOnly timing_noise_audit is issued (autolens_profiling#362: open, no comments, unassigned); start its audit phase via start_dev. Other nine tasks unchanged and unissued; mass_field superseded confirmed. Active tasksAudit timing tests and profiling gates for measurement noise↗Issued as autolens_profiling#362 (open, 0 comments, unassigned). Start its read-and-classify audit phase via start_dev. MGE likelihood_breakdown steps are cumulative and `linear_gaussians` is reported as 0Unchanged and unissued (no GitHub refs, 2026-10-04). Select one bounded step when measurement-tools is prioritised. A gradient-cost probe: forward vs `value_and_grad` ms/eval and a strict FD check, on any registry cellUnchanged and unissued (workspace_developer#117 closed 2026-07-28). Select one bounded step when prioritised. Numba breakdown harness: perturb the instance so the operated-matrix memo cannot hide a stepUnchanged and unissued (PyAutoArray#496 closed 2026-08-27). Select one bounded step when prioritised. Search settings-estimation + profiling infrastructure (n_starts / batch_size / n_batch)Unchanged and unissued (autolens_profiling#82 closed 2026-08-18). Select one bounded step when prioritised. jax_compile/probe.py lost its cell builder with the searches tier — give profiling its ownUnchanged and unissued; scripts/misc/jax_compile/probe.py still on main (run state unverified). Select one bounded step when prioritised. profile_lens_aggregator.py cannot run from the autolens_workspace_developer root: no config/ directoryUnchanged: workspace_developer main still has no root config/ (bug presumed live, not re-run). Select one bounded step when prioritised. jax_profiling/gradient/imaging/pixelization.py: 3.2% of its pin move is unattributedUnchanged and unissued (PyAutoArray#490 merged 2026-08-26). Select one bounded step when prioritised. Pair JAX/XLA env vars with measured compile and run times, per backendUnchanged and unissued. Context: jax-version policy changes merged to library mains 2026-10-04, unreleased. Select one bounded step when prioritised. | active | |
PyAutoFit profiling bootstrap1 task · next stepPyAutoLabs/autofit_profiling does not exist (gh, 2026-10-04). Human decision pending on repo creation; no producer registration until it exists with ported baselines. Active tasksautofit_profiling: bootstrap the repo + general PyAutoFit profiling epicHuman answers the single repo-creation question (PyAutoLabs/autofit_profiling absent, 2026-10-04) before any porting. | needs-decision | |
Critical curves and evaluation grids0 tasks · next stepRelease gate met: PyAutoGalaxy#646 first released in 2026.10.4.1. 2026-10-04: /prm of autolens_workspace_test#343 was denied by the auto-mode classifier ("Merge Without Review"); the human runs it outside auto mode to close the evaluation-grid-cap-field task. Active tasksNo open tasks. | needs-decision | |
Certified positive solver0 tasks · next stepParked by choice; wiki unchanged since ab1e4fd (2026-09-27). Its three library/workspace Mind drafts remain in Mind. Retain parked; read the project ledger before proposing new work. Active tasksNo open tasks. | parked |
Choose a project, dataset and model. Qualification belongs to each recorded setup.
Project autolens_profiling, scope setup-catalogue, read from PyAutoLabs/autolens_profiling dashboard/catalogue.json at 4832679a; producer revision a93f37a7; generated 2026-10-04T18:28:06Z; policy no-temporal-comparisons. Project dashboard · receipt.
Project autolens_profiling, scope setup-catalogue, read from PyAutoLabs/autolens_profiling dashboard/catalogue.json at 4832679a; producer revision a93f37a7; generated 2026-10-04T18:28:06Z; policy no-temporal-comparisons. Project dashboard · receipt.