Repository navigation
cortex: activate ic50_workspace and submit graphical-ep phase 4 - #26
Merged
Merged
Conversation
Phase 4 of the graphical-ep campaign (IC50 use case) is live on RAL.
projects.yaml — the row described a project it could not actually read. Three
fields were wrong, all now verified against the checkout and the cluster:
* sync_verbs: [pull] -> the real 17-verb list. hpc/sync is a 560-line CLI with
push, submit and push-submit implemented; the row understated it so far that
the project looked unable to reach RAL at all.
* ledger: README.md -> NOTES.md. ic50_workspace has no README.md; NOTES.md is
its commentary ledger. Flagged in the note: NOTES.md is dated 2026-05-07 and
predates the graphical-fit, HPC and profiling work, so it needs a refresh.
* status: dormant -> active, partition: both -> ral. The blockers recorded on
2026-09-08 (dirty checkout, no results/ witness path) are cleared: the
checkout is clean, results/ exists at the project root and matches the
witness glob, and /mnt/ral/jnightin/ic50_workspace now exists and has run.
tasks/ic50_workspace/ep_scale_up.md — planned -> ready -> submitted, runs 342408
(EP, sim, nlive 150, max_steps 12) and 342409 (graphical, sim, nlive 150), both
on the ral partition. The task's "Ready when" line still told a reader to gate it
on phase 2, which shipped on 2026-09-02; replaced with what actually happened,
including that the RAL PyAutoFit mirror was checked for #1580 before submitting.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PMovQR9KqQUVBEg2b3EAsX
The two generated pages are ledger for the merge gate, and dashboard_refresh.yml errors rather than self-heals on a PR run, so the board has to be regenerated in the branch. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PMovQR9KqQUVBEg2b3EAsX
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Phase 4 of the graphical-ep campaign (the IC50 use case) is now running on RAL. This brings the Cortex row and task into line with reality.
Why the row needed fixing
The
ic50_workspacerow described a project it could not actually read. Three fields were wrong, all now verified against the checkout and the cluster:sync_verbs[pull]hpc/syncis a 560-line CLI withpush,submit,push-submitimplemented; the row understated it so far the project looked unable to reach RAL at allledgerREADME.mdNOTES.mdic50_workspacehas noREADME.mdstatusdormantactivepartitionbothralforce_x1_cpu=True, no pool) andgpuis down to one usable nodeThe recorded blockers are cleared
The ledger parked this task on 2026-09-08 because the checkout was dirty and there was no
results/witness path. Both are resolved:core.fileMode=false), the rest regenerated artifacts. Tree is clean.results/now exists at the project root and matches the row'switness_file: results/**/*.jsonglob. It never did before: the scripts wrote toscripts/results/, which the glob could not match, andPULL_DIRS=(output)would never have retrieved it from RAL either. Both fixed./mnt/ral/jnightin/ic50_workspacedid not exist. It does now, and has run.Runs
tasks/ic50_workspace/ep_scale_up.md:planned → ready → submitted.nlive 150,max_steps 12, partitionralnlive 150, partitionralSim-first by design:
dataset/ic50_sim/carries per-datasettrue_params.jsonplus_sample/ground_truth.json, so recovery is checkable against truth.The RAL PyAutoFit mirror was verified to contain
08207bad0(#1580, the EP stale-mask fixed-point fix) before submitting — so unlike phase 3's EP arm, these runs carry the complete D1–D6 fix wave.One caveat recorded in the note
NOTES.mdis dated 2026-05-07 and predates the graphical-fit, HPC and profiling work — its architecture diagram does not mentiongraphical_*.pyorhpc/at all. It is the right ledger file, but it needs a refresh. Recorded in the row'snote:rather than silently accepted.cortex.py check: OK.ledger_merge.py classifycorrectly routes this to a human becauseprojects.yamlis code.🤖 Generated with Claude Code
https://claude.ai/code/session_01PMovQR9KqQUVBEg2b3EAsX