Skip to content

campaigns: hst_mge_cpu_pilot superseded by the phase-2a MGE/HST CPU cell - #75

Merged
Jammy2211 merged 3 commits into
mainfrom
feature/likelihood-lifecycle-p2a-pulse-adopt
Oct 25, 2026
Merged

Jammy2211 merged 3 commits into
mainfrom
feature/likelihood-lifecycle-p2a-pulse-adopt

Conversation

@Jammy2211

Copy link
Copy Markdown
Contributor

Summary

Closes the hst_mge_cpu_pilot task (setup-baseline campaign, autolens_profiling#415) as superseded. By human ruling on 2026-10-25, the pilot is answered by the phase-2a imaging campaign's MGE/HST CPU cell: imaging/mge/hst jax_cpu_dense_fp64@hpc_ral_cpu, measured at tag 2026.10.23.1 with parity match, in autolens_profiling#461.

The pilot's own protocol was not run. tasks/hst_mge_cpu_pilot.md records how the cell differs from it:

  • production MGE at the seeded prior draw, rather than the 20+20 Gaussian instance;
  • single-jit medians and vmap per-call times, rather than cold / warm / persistent-cache stages;
  • no background-load qualification;
  • a per-cell parity witness, rather than the operator + scipy NNLS witness.

The task file also lists which of the pilot's questions stay open. Those questions get a new task if they matter again.

Merge order: autolens_profiling#461 merges first. This branch then refetches the lens receipt, so Pulse ingests the new @3 records (104 rows on 2026.10.23.1), and re-renders the board before merging. As it stands the board is rendered offline from the committed receipts.

Scripts Changed

  • campaigns.yaml — hst_mge_cpu_pilot: status superseded, next, evidence, updated
  • tasks/hst_mge_cpu_pilot.md — status line and a "Superseded by phase 2a" section
  • dashboard.md, dashboard.html — re-rendered (pyauto-pulse board --offline)

Upstream PR

PyAutoLabs/autolens_profiling#461

Test Plan

  • pyauto-pulse check --offline OK; pytest tests 296 passed; ruff check + format clean
  • After #461 merges: pyauto-pulse board online (lens receipt refetch), then check, then commit

Refs PyAutoLabs/autolens_profiling#456, PyAutoLabs/autolens_profiling#415

Generated by the PyAutoLabs agent workflow.

🤖 Generated with Claude Code

https://claude.ai/code/session_014iaN7eDPr9toV6UcpDKTuZ

Jammy2211 and others added 3 commits October 25, 2026 16:37
…ell (#461)

Human ruling 2026-10-25: adopt the likelihood-lifecycle phase-2a imaging
cell imaging/mge/hst jax_cpu_dense_fp64@hpc_ral_cpu (tag 2026.10.23.1,
parity match, autolens_profiling#461) as the pilot's answer and close the
pilot. Its own protocol was not run; the task file records how the cell
differs and which pilot questions stay unanswered.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014iaN7eDPr9toV6UcpDKTuZ
… at eef5edfe)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014iaN7eDPr9toV6UcpDKTuZ
@Jammy2211
Jammy2211 merged commit cf46a5a into main Oct 25, 2026
2 checks passed
@Jammy2211
Jammy2211 deleted the feature/likelihood-lifecycle-p2a-pulse-adopt branch October 25, 2026 16:51
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant