PyAutoCortexDashboard

subhalo_validation — Do the SLaM subhalo pipelines false-detect on lenses with no subhalo

active · ral · subhalo_validation#2 · projects/subhalo_validation.md · local /mnt/c/Users/Jammy/Science/subhalo_validation · RAL /mnt/ral/jnightin/subhalo_validation

Now

Five chains on RAL under the corrected adapt settings: rectangular_adapt on pl_eff_0 / pl_eff_1_outer (job A done, job B 342311 submitted) and the delaunay_adapt_split reruns plus rectangular_adapt on pl_sersic_0. pl_sersic_0 on delaunay_adapt_split is accepted (no false detection, +2.557).
Next: pull when the numba chains land and read the evidence_increase of each; no more 1000-pixel runs.

Runs

• 342299_1 — open — ral — 2026-09-07 — delaunay_adapt_split_pl_eff_0: job B, numba chain at the standard 1250-pixel Delaunay mesh (8c / 96gb / 48 h); reloads job A 342231_1; the 09-07 ruling: run al…
• 342299_2 — open — ral — 2026-09-07 — delaunay_adapt_split_pl_eff_1_outer: job B, numba chain at the standard 1250-pixel Delaunay mesh (8c / 96gb / 48 h); reloads job A 342231_2; the 09-07 ruling:…
• 342311_1 — open — ral — 2026-09-07 — rectangular_adapt_pl_eff_0: job B, numba chain (8c / 96gb / 48 h) at the standard 1250-pixel mesh under PIPELINE=rectangular_adapt; reloads job A 342237_1; the…
• 342311_2 — open — ral — 2026-09-07 — rectangular_adapt_pl_eff_1_outer: job B, numba chain (8c / 96gb / 48 h) at the standard 1250-pixel mesh under PIPELINE=rectangular_adapt; reloads job A 342237_…
• 342311_0 — open — ral — 2026-09-07 — rectangular_adapt_pl_sersic_0: job B, numba chain (8c / 96gb / 48 h) at the standard 1250-pixel mesh under PIPELINE=rectangular_adapt; reloads job A 342237_0;…

Last 5

• 2026-09-07 — result — accepted delaunay_adapt_split_pl_sersic_0 (R-20260907-01): Accepted. The numba delaunay_adapt_split chain on pl_sersic_0 (job 342273_0, 22 h 08 m, 1250 Hilbert / 1310 Delaunay pixels) lands evidence_increase +2.557 (< 5): no false detectio…
• 2026-09-07 — run — 342311_0 submitted: rectangular_adapt_pl_sersic_0 — job B, numba chain (8c / 96gb / 48 h) at the standard 1250-pixel mesh under PIPELINE=rectangular_adapt; reloads job A 342237_0; the 09-07 ask: submit the rectangular runs too
• 2026-09-07 — run — 342311_2 submitted: rectangular_adapt_pl_eff_1_outer — job B, numba chain (8c / 96gb / 48 h) at the standard 1250-pixel mesh under PIPELINE=rectangular_adapt; reloads job A 342237_2; the 09-07 ask: submit the rectangular runs too
• 2026-09-07 — run — 342311_1 submitted: rectangular_adapt_pl_eff_0 — job B, numba chain (8c / 96gb / 48 h) at the standard 1250-pixel mesh under PIPELINE=rectangular_adapt; reloads job A 342237_1; the 09-07 ask: submit the rectangular runs too
• 2026-09-07 — run — 342299_2 submitted: delaunay_adapt_split_pl_eff_1_outer — job B, numba chain at the standard 1250-pixel Delaunay mesh (8c / 96gb / 48 h); reloads job A 342231_2; the 09-07 ruling: run all datasets at 1250, no more pix1000

full log

euclid_dr1_prelim — Refit the ten DR1-prelim tiles and rebuild the catalogue on RAL

active · both · PyAutoCortex#35 · projects/euclid_dr1_prelim.md · local /mnt/c/Users/Jammy/Science/euclid_dr1_prelim · RAL /mnt/ral/jnightin/euclid_dr1_prelim

Now

342629 reran the ten tiles on refreshed libraries; eight finished and confirmed all four 09-10 defects fixed, tiles 1 and 3 are still in vis_lp. The SED chain ran on GPU for the eight (342648) and the CPU verification on tile 9 (342668_9) passed in 9:24, faster than the GPU — the CPU route is the one the main experiment uses.
Next: pull, bundle with output_sed, compare magnitudes (µJy fluxes) against the 20260623 catalogue; then tiles 1/3 SED on the CPU script. The parity clause needs restating: rerun scatter, not ordering, is what fails it.

Runs

• 342629_[0-9] — running — ral — 2026-09-10 — ordered_rerun_catalogue_parity: two-stage vis_lp(n_live=750)+vis_pix(n_live=300), array 0-9, ten dr1_prelim tiles; libraries refreshed BEFORE submit (PyAutoNer…
• 342648_[0,2,4-9] — open — gpu — 2026-09-11 — ordered_rerun_catalogue_parity: SED chain `hpc/batch_gpu/submit_sersic_waveband_8tiles` (a copy of `submit_sersic_waveband` differing only in `--array=0,2,4-9`…
• 342668_9 — open — ral — 2026-09-11 — ordered_rerun_catalogue_parity: CPU verification of the SED chain via hpc/batch_cpu/submit_sersic_waveband (pipeline PR #70, sbatch --array=9): tile 102008848,…
• 342796_[1,3,8] — open — ral — 2026-09-12 — ordered_rerun_catalogue_parity: CPU SED chain for the three tiles still missing one, via a direct ssh sbatch --array=1,3,8 on hpc/batch_cpu/submit_sersic_waveb…

Last 5

• 2026-09-12 — run — 342796_[1,3,8] submitted: ordered_rerun_catalogue_parity: CPU SED chain for the three tiles still missing one, via a direct ssh sbatch --array=1,3,8 on hpc/batch_cpu/submit_sersic_waveband (md5 65db7d8b… on RAL = the clone's commit 88dda9d…
• 2026-09-11 — note — **CPU verification PASSED and was faster than the GPU.** `342668_9` (ral, 8 CPU, JAX on the CPU backend) COMPLETED in **9:24** wall, exit 0, all eight `sersic_lens_model/{vis,nir_y,nir_j,nir_h,decam_g,decam_r,decam_i,decam_z}` searches wri…
• 2026-09-11 — note — **342629 scored on the eight tiles that finished; all four 09-10 defects confirmed fixed; parity fails on rerun scatter, not ordering.** `sacct`: tasks 0,2,4,5,6,7,8,9 COMPLETED (walls 5:45–9:42, MaxRSS 7.6–10.9 GB); tasks 1 (`Tile10200729…
• 2026-09-11 — run — 342668_9 submitted: ordered_rerun_catalogue_parity — CPU verification of the SED chain via hpc/batch_cpu/submit_sersic_waveband (pipeline PR #70, sbatch --array=9): tile 102008848, JAX on the CPU backend, 8 CPU; same seeded vis_lp zip as t…
• 2026-09-11 — run — 342648_[0,2,4-9] submitted: ordered_rerun_catalogue_parity — SED chain `hpc/batch_gpu/submit_sersic_waveband_8tiles` (a copy of `submit_sersic_waveband` differing only in `--array=0,2,4-9`, because `hpc/sync submit` passes no sbatch argume…

full log

analytic_gaussian — Graphical and EP against a closed-form Gaussian posterior, at ensemble scale

active · ral · PyAutoCortex#34 · projects/analytic_gaussian.md · local /mnt/c/Users/Jammy/Science/analytic_gaussian · RAL /mnt/ral/jnightin/analytic_gaussian

Now

Wave 1 (342413, 200 seeds at N=5) is accepted as the baseline: autofit EP is exact on the Gaussian leg, leg B sigma misses as pre-registered (78/200), collapse rate 0/200. Follow-ups filed through /intake.
Next: settle whether criterion 2's mu threshold or the minimal-EP control is wrong before any rerun; the N=25 rung is written but not submitted.

Runs

nothing on the cluster

Last 5

• 2026-09-10 — result — accepted ensemble_parity (R-20260910-04): ok accept and intake the things to address. we can do another run down the line so also achieve the results in the analytic_gaussian project for future comparison
• 2026-09-10 — note — question: Is criterion 2's mu threshold or the minimal-EP control wrong — planned, never run [archive/tasks/analytic_gaussian/minimal_ep_legb_mu_threshold.md]
• 2026-09-09 — run — 342413_[0-199] submitted (done, wall 0:05): ensemble_parity — the N=5 seed ensemble, 200 seeds, 50 concurrent, `hpc/batch_cpu/submit_ensemble_n5`, sample `ens_n5`. RAL PyAuto mirror verified at PyAutoFit `66f9f8d5d` before submission, whic…
• 2026-09-09 — note — question: Do graphical and EP recover closed-form means and errors [archive/tasks/analytic_gaussian/ensemble_parity.md]

full log

ep_toy_gaussian — NUTS versus three EP fits on the #1405 collapse toy at N=50

active · ral · ep_toy_gaussian#1 · projects/ep_toy_gaussian.md · local /mnt/c/Users/Jammy/Science/ep_toy_gaussian · RAL /mnt/ral/jnightin/ep_toy_gaussian

Now

n50_seed42 is on RAL: NUTS + one-by-one as 342639, EP x3 on identical data as 342640_[0-2]. The one-by-one leg finished (50/50 in 319 s); the EP repeats are turning per-factor Dynesty fits.
Next: pull and run scripts/compare.py --sample n50_seed42; settle first whether a prior factor with only NO_CHANGE rows counts as zero SUCCESS updates.

Runs

• 342639 — open — ral — 2026-09-10 — nuts_vs_ep_x3_n50: one_by_one + BlackJAX NUTS, partition ral
• 342640 — open — ral — 2026-09-10 — nuts_vs_ep_x3_n50: EP x3 on identical data, array 0-2

Last 5

• 2026-09-10 — run — 342640 submitted: nuts_vs_ep_x3_n50 — EP x3 on identical data, array 0-2
• 2026-09-10 — run — 342639 submitted: nuts_vs_ep_x3_n50 — one_by_one + BlackJAX NUTS, partition ral
• 2026-09-10 — note — question: Do three EP fits agree with NUTS at N=50 [archive/tasks/ep_toy_gaussian/nuts_vs_ep_x3_n50.md]

full log

slope_hierarchy_scale — Hierarchical slope recovery with graphical and EP at N=25 to 50

active · both · slope_hierarchy_scale#2 · projects/slope_hierarchy_scale.md · local /mnt/c/Users/Jammy/Science/slope_hierarchy_scale · RAL /mnt/ral/jnightin/slope_hierarchy_scale

Now

The N=25 scale-up is running: graphical array 342348_[0-24], joint fits 342350_0 / 342351_0, and the EP rerun 342410 on the mirror that carries #1580.
Next: pull wave 1 and compare EP against the NUTS parent sigma; then decide whether to go to N=50.

Runs

• 342348_[0-24] — open — gpu — 2026-09-08 — n25_scale_up: one lens per array task, hpc/batch_gpu/submit_one_by_one, sample_n25_seed42
• 342350_0 — open — gpu — 2026-09-08 — n25_scale_up: joint hierarchical NUTS fit, hpc/batch_gpu/submit_graphical
• 342351_0 — open — ral — 2026-09-08 — n25_scale_up: EP arm on the CPU partition, hpc/batch_cpu/submit_ep, max_steps 12; RAL PyAutoFit mirror 68ff9bd57 predates PyAutoFit#1580, so this arm runs with…
• 342410 — open — ral — 2026-09-09 — n25_scale_up: EP arm rerun on the JAX-on-CPU path (no multiprocessing pool); replaces 342351_0

Last 5

• 2026-09-09 — run — 342410 submitted: n25_scale_up — EP arm rerun on the JAX-on-CPU path (no multiprocessing pool); replaces 342351_0
• 2026-09-08 — run — 342351_0 submitted: n25_scale_up — EP arm on the CPU partition, hpc/batch_cpu/submit_ep, max_steps 12; RAL PyAutoFit mirror 68ff9bd57 predates PyAutoFit#1580, so this arm runs without the stale-mask fixed-point fix
• 2026-09-08 — run — 342350_0 submitted: n25_scale_up — joint hierarchical NUTS fit, hpc/batch_gpu/submit_graphical
• 2026-09-08 — run — 342348_[0-24] submitted: n25_scale_up — one lens per array task, hpc/batch_gpu/submit_one_by_one, sample_n25_seed42
• 2026-07-22 — note — question: Does hierarchical slope recovery hold at N=25 to 50 [archive/tasks/slope_hierarchy_scale/n25_scale_up.md]

full log

ic50_workspace — EP against the graphical joint fit on IC50 dose-response data, scaling up

active · ral · PyAutoCortex#36 · projects/ic50_workspace.md · local /mnt/c/Users/Jammy/Science/ic50_workspace · RAL /mnt/ral/jnightin/ic50_workspace

Now

Parity at N=5 achieved on RAL (33/33 within 3σ for both methods, runs 342408 / 342409); the scale ladder runs 342411 and 342412 are submitted.
Next: pull the ladder and see where EP's cost and hill_coef width go as N grows; ep_lbfgs_jax is the scale lever.

Runs

• 342408 — open — ral — 2026-09-09 — ep_scale_up: EP sim, nlive 150, max_steps 12
• 342409 — open — ral — 2026-09-09 — ep_scale_up: graphical sim joint Dynesty, nlive 150
• 342411 — open — ral — 2026-09-09 — ep_scale_up: EP scale ladder, sim rungs 5/10/25/50, nlive 150 max_steps 12
• 342412 — open — ral — 2026-09-09 — ep_scale_up: graphical scale ladder, sim rungs 5/10/25/50, nlive 150

Last 5

• 2026-09-09 — run — 342412 submitted: ep_scale_up — graphical scale ladder, sim rungs 5/10/25/50, nlive 150
• 2026-09-09 — run — 342411 submitted: ep_scale_up — EP scale ladder, sim rungs 5/10/25/50, nlive 150 max_steps 12
• 2026-09-09 — run — 342409 submitted: ep_scale_up — graphical sim joint Dynesty, nlive 150
• 2026-09-09 — run — 342408 submitted: ep_scale_up — EP sim, nlive 150, max_steps 12
• 2026-08-19 — note — question: Does EP match the graphical joint fit end to end [archive/tasks/ic50_workspace/ep_scale_up.md]

full log

autolens_inference — Do the PyAutoLens inference backends agree on the HST SLaM posterior

active · both · autolens_inference#4 · projects/autolens_inference.md · local /home/jammy/Code/PyAutoLabs/autolens_inference · RAL /mnt/ral/jnightin/autolens_inference

Now

The four A100 base legs are done (dense and sparse, seeds 0-1, all five stages): sparse costs 1.5x dense and buys nothing measurable, and at the same seed the two routes agree to 0.0-0.6 sigma while seeds of the SAME leg disagree by up to 0.9 sigma and ~3 nats — so the slam_hst_base witness bands (2 nats, 0.2 sigma) are tighter than Nautilus's own seed reproducibility on this cell, which is a finding about the bands. Running now: the same chain with a 1250-vertex Hilbert/Delaunay source (343143 dense, 343145 sparse, seeds 0-1) on the A100. Next: pull both, compare seed-paired against slam_base, and decide whether the parity bands need rewriting before the four CPU legs are submitted.

Runs

• 343143_[0-1] — open — gpu — 2026-09-15 — slam_hst_delaunay_1250: A100 leg, dense inversion, seeds 0-1, hpc/batch_gpu/submit_slam_delaunay1250_hst_jax_gpu_dense (config hpc_a100_jax_gpu_dense_fp64, var…
• 343145_[0-1] — open — gpu — 2026-09-15 — slam_hst_delaunay_1250: A100 leg, sparse inversion, seeds 0-1, hpc/batch_gpu/submit_slam_delaunay1250_hst_jax_gpu_sparse (config hpc_a100_jax_gpu_sparse_fp64,…

Last 5

• 2026-09-15 — run — 343145_[0-1] submitted: slam_hst_delaunay_1250: A100 leg, sparse inversion, seeds 0-1, hpc/batch_gpu/submit_slam_delaunay1250_hst_jax_gpu_sparse (config hpc_a100_jax_gpu_sparse_fp64, variant delaunay_1250) — same mesh, sparse operator route
• 2026-09-15 — run — 343143_[0-1] submitted: slam_hst_delaunay_1250: A100 leg, dense inversion, seeds 0-1, hpc/batch_gpu/submit_slam_delaunay1250_hst_jax_gpu_dense (config hpc_a100_jax_gpu_dense_fp64, variant delaunay_1250) — 1250-vertex Hilbert/Delaunay sourc…
• 2026-09-14 — result — "the first runs looks good" (human, 2026-09-14) — the four A100 base legs, dense and sparse at seeds 0 and 1, all five stages complete. Next, on the same ask: the same chain with a 1250-vertex Delaunay source instead of the 28x28 rectangul…
• 2026-09-14 — run — 342801_[0-1] finished — wall 1:07: slam_hst_base: A100 leg, sparse inversion, seeds 0-1, hpc/batch_gpu/submit_slam_hst_jax_gpu_sparse (config hpc_a100_jax_gpu_sparse_fp64); the 09-12 ask: one hardware type first, GPU; rate-probe 342695 tre…
• 2026-09-14 — run — 342799_[0-1] finished — wall 0:48: slam_hst_base: A100 leg, dense inversion, seeds 0-1, hpc/batch_gpu/submit_slam_hst_jax_gpu_dense (config hpc_a100_jax_gpu_dense_fp64); the 09-12 ask: one hardware type first, GPU; rate-probe 342695 tree m…

full log

euclid — The original Euclid DR1 grade-AB catalogue project, superseded by euclid_dr1_prelim

dormant · both · no issue yet · projects/euclid.md · local /mnt/c/Users/Jammy/Science/euclid · RAL /mnt/ral/jnightin/euclid_strong_lens_modeling_pipeline

Now

Dormant since 2026-09-07: superseded by euclid_dr1_prelim. Three questions were planned and never run (see the archive).

Runs

nothing on the cluster

Last 5

• 2026-08-28 — note — question: Do we recover Sersic indices or is prior pile-up artefact — planned, never run [archive/tasks/euclid/sersic_index_recovery.md]
• 2026-08-28 — note — question: Can a fitted Euclid lens be resimulated and recovered — planned, never run [archive/tasks/euclid/resimulate_fitted_lens_simulator.md]
• 2026-08-28 — note — question: How robust are magnification estimates under model match and mismatch — planned, never run [archive/tasks/euclid/magnification_robustness.md]

full log