task #1093: verify_plan check c27 — 7B activation capture booked on eval/debug intent#858
Merged
Conversation
…val/debug (L4) intent Mechanizes plan-compute-sizing § Activation-capture HBM sizing (the #666/#744 OOM class; founding false negative #825 v17). FAIL (experiment) / WARN (analysis), conditional; reuses c26's intent machinery; intent sets derived from the _C26_INTENT_GPU mirror (BIG by exclusion); four permissive skips (RunPod pin, pod.py provision, window H100/H200, big-intent absolution); N/A escape 'N/A — no 7B activation capture'. Scale-qualified, self-DISARM-safe FAIL detail (no H100/H200/backend-runpod tokens). Docstring: catalog row, conditional enumeration + N/A phrase list. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…tion-pin updates Threshold-semantics matrix (12b), mirror-partition assert (15), self-DISARM cells (16/16b), vocab/skip-arm matrices (17), kind gates, N/A anti-paste, #522 wrapped-window unit test. Updates test_good_plan_passes_all (c27 SKIP row, 26→27) and the CLI schema pins (27→28 checks, n_skip 19→20). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…paragraph Coverage-honest wording: c27 backstops GCP/auto-lane eval/debug bookings only; RunPod-pinned plans + phase-to-intent routing stay critic-owned. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
superkaiba
marked this pull request as ready for review
July 7, 2026 04:59
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes task #1093 (workflow-fix, kind: infra).
Summary
check_capture_intent_hbm) toscripts/verify_plan.py: FAIL (experiment) / WARN (analysis) when a plan carries activation-capture vocabulary + a ≥7B model signal while booking theeval/debug(GCP L4, 16-GB-class) intent — the issue-868: extraction.py docstring — last-layer final-RMSNorm caveat + PEFT fallback trap (doc-only) #666/issue-976: widen runpod stale-sentinel clear to experimenter step-11 wildcard #744 OOM class; founding false negative task #1056: critics verify figures against pinned blob (pin-first rule) #825 plan v17 now FAILs, v18 PASSes, issue-976: widen runpod stale-sentinel clear to experimenter step-11 wildcard #744 v2 FAILs.tests/test_verify_plan.py(incl. threshold-semantics size regex, partition assert, self-DISARM asserts, registration-pin updates); 592-newest-plan corpus re-scan: 1 genuine flag, 0 false positives..claude/rules/plan-compute-sizing.md.Test plan
uv run pytest tests/test_verify_plan.py -q— 440 passedstep9c_baseline.py compareexit 0 (no new failures, no ruff regression)🤖 Generated with Claude Code