docs(autotrain): continuous-openui-local screening cycles 11-29 closeout - #1512
Merged
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Warning Review limit reached
Next review available in: 48 minutes You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (41)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Tyler-R-Kendrick
changed the title
4 tasks
Tyler-R-Kendrick
changed the title
Cycle continuous-loop-20260808-continuous-openui-local-8c0b60dd-c11 (component-edge vs control, wf_smoke_v2 fixture): non-positive fixture screen, still fails honest ship gates. Local docs commit only, no stacked PR. Appends the matching no-bump version-stamp history entry for harness.experiments.slm228_spectral_disposition. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QBCuFmvGEyYJsjKGCccbNQ
…0260808-continuous-openui-local-8c0b60dd-c12 closeout
…0260808-continuous-openui-local-8c0b60dd-c13 closeout
…0260808-continuous-openui-local-8c0b60dd-c14 closeout
…0260808-continuous-openui-local-8c0b60dd-c15 closeout
…0260808-continuous-openui-local-8c0b60dd-c16 closeout
…0260808-continuous-openui-local-8c0b60dd-c17 closeout
…0260808-continuous-openui-local-8c0b60dd-c18 closeout
…0260808-continuous-openui-local-8c0b60dd-c19 closeout
…0260808-continuous-openui-local-8c0b60dd-c20 closeout
…0260808-continuous-openui-local-8c0b60dd-c21 closeout
…0260808-continuous-openui-local-8c0b60dd-c22 closeout
…0260808-continuous-openui-local-8c0b60dd-c23 closeout
…0260808-continuous-openui-local-8c0b60dd-c24 closeout
…0260808-continuous-openui-local-8c0b60dd-c26 closeout
…0260808-continuous-openui-local-8c0b60dd-c27 closeout
…0260808-continuous-openui-local-8c0b60dd-c28 closeout
…0260808-continuous-openui-local-8c0b60dd-c29 closeout
Cycle continuous-loop-20260808-continuous-openui-local-8c0b60dd-c25
completed training+eval (campaign.json, results.tsv, events.jsonl,
runs/*/scoreboard.json all present) but never got a documented
closeout: _self_heal_document_actions() requires cycle_handoff.json
to exist and no-ops without it (scripts/run_autotrain_continuous.py:
2570-2583), and that file was never written for c25 before the
supervisor moved on to c26 in the next invocation. The cycle's docs
were silently skipped, violating the repo's Iron Law (docs follow
every experiment).
Backfills docs/design/*-c25-results.{md,json} directly from the
committed run scoreboards (both arms: structural_similarity=0.3764,
null delta, non-positive/gate-reject at n=3 fixture scale — consistent
with every other cycle this session), the matching README/MODEL_CARD
checkpoint-provenance note, and a no-bump version-stamp entry. Flags
the underlying driver gap in both the results doc and the version-stamp
note as a harness signal for a future repair_harness follow-up.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QBCuFmvGEyYJsjKGCccbNQ
Tyler-R-Kendrick
force-pushed
the
claude/great-dirac-rx8wy9
branch
from
August 9, 2026 04:06
eeb1c38 to
d66235d
Compare
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
2 participants
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.
Summary
Third
/autotraincontinuous-loop iteration of this scheduled run (follow-upto #1480 and #1509, both merged). Restarted the designated branch from the
latest
mainper the merged-PR restart rule and ran 18 more supervisedscreening cycles of
continuous-openui-local(wf_smoke_v2fixture traindata, 20 steps, honest ship gates on), landing incrementally as this branch
stayed open:
docs/design/continuous-loop-20260808-continuous-openui-local-8c0b60dd-c{11..29}-results.{md,json}— one JSON + markdown pair per cycle (
documenting-experiment-results).docs/MODEL_CARD.md/README.md— per-cycle checkpoint-provenancenotes (fixture/scratch, not a ship promotion), each heading suffixed with
its campaign id per the MD024 convention landed in docs(autotrain): continuous-openui-local screening cycles 1-5 closeout #1480.
src/slm_training/resources/versions.json— matchingno-bump:historyentries for
harness.experiments.slm228_spectral_disposition.Honesty status — all 19 cycles are non-positive
Every cycle correctly failed the honest fixture ship gates (expected at
n=3fixture scale) or hit a soft, self-healing infrastructure signal (armnot executed this pass, frozen-replay retry). Per
autotrain'scontinuous-mode law, non-positive cycles get local docs commits only — no
gh stacklayer, so this PR is plain documentation/provenancehousekeeping, not a claimed model or harness win. No checkpoint here is
being promoted or shipped.
Fixed: cycle c25's missing docs closeout (Iron Law gap)
Cycle c25 completed training+eval (its
campaign.json,results.tsv,events.jsonl, and both arms'scoreboard.jsonare present) but the drivernever wrote its
cycle_handoff.jsonbefore the next invocation moved on toc26 —
_self_heal_document_actions()requires that file and silentlyno-ops without it
(
scripts/run_autotrain_continuous.py:2570-2583), so c25's results wereskipped rather than documented. That's an Iron Law violation ("docs follow
every experiment"), caught by manual audit of the run matrix vs. committed
docs. Backfilled
docs/design/*-c25-results.{md,json}directly from thecommitted run scoreboards (both arms:
structural_similarity=0.3764, nulldelta — consistent with every other cycle this session) and flagged the
underlying driver gap as a harness signal worth a future
repair_harnessfollow-up.
Test plan
scripts.verify_version_stamps --checkpasses on every commit in this PRtests/test_scripts/test_verify_checkpoint_references.py,tests/test_versioning/*) pass on every commitdocs/design/*-results.json(including c25, reconstructed from thecommitted scoreboards)
interpret any of these 19 cycles as ship-ready