pgw#1242: cut 0.114.3 — resume_from, the fifth reserved model input - #794
Merged
Conversation
Cut from the PINNED commit d9174a1; 31 commits since v0.114.2, assembled from 7 fragments, under an enqueue freeze with the queue verified empty and all seven open PRs pinned and --auto-disarmed. WHY NOW: a funded H100 booking needs `resume_from` (pgw#1242 / te#185) published. This wheel is NOT "the mint is fixed". WHAT IS NOT IN IT: pgw#1243's finalize-wedge fix. It is PR #791, still open and UNMERGEABLE in the queue when this was pinned, and it is the P0 standing between the platform and its first published compiled graph. Read this wheel as everything EXCEPT the wedge; the fix gets 0.114.4. Also rides the th#1834 Phase 3 mint supervision cut (pgw#1215: the serving parent supervises its own compile children and the middle mint tier is deleted), pgw#1206 C3, pgw#1237, pgw#1239, th#1887 and th#1918. THE SWEEP FOUND th#1918 OWED A FRAGMENT AND HAD NONE — the .yml -> .yaml workflow rename, which is exactly the entry a reader will want when this release's publish 403s. Written rather than shipped silent: PyPI trusted publishing binds to the workflow FILENAME, 0.114.2 published on the old name an hour before the rename landed, so THIS is the first tag through publish.yaml. If the human-side PyPI edit has not landed the upload 403s after CI, build and provenance pass — uploading nothing, invalidating nothing, and re-runnable on the same tag.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Cut from pinned
d9174a18. 31 commits sincev0.114.2, 7 fragments. Under an enqueue freeze (pgw#1226) with the queue verified empty and all seven open PRs pinned and--auto-disarmed at freeze time.Why now: a funded H100 booking needs
resume_from(pgw#1242 / te#185) published.⛔ What is NOT in this wheel
pgw#1243's finalize-wedge fix does not ride. It is PR #791, still open — and
UNMERGEABLEin the queue — when this was pinned. That wedge is the P0 standing between the platform and its first published compiled graph, so read 0.114.3 as "everything except the wedge". The fix gets 0.114.4.Contents
resume_fromas the fifth reserved model input (pgw#1242) · the th#1834 Phase 3 mint supervision cut — the serving parent supervises its own compile children and the middle mint tier is deleted (pgw#1215, four PRs) · pgw#1206 C3 residency teardown · pgw#1237 worker parity · pgw#1239 Hub-worker launch fence · th#1887 env deletion · th#1918 workflow rename.publish.yaml— a 403 is expected, not a defectth#1918 (PR #784) renamed the workflows at 02:47Z; 0.114.2 published at 01:44Z on the old
publish.yml. PyPI trusted publishing binds to the workflow filename, so unless the human-side publisher edit has landed, the upload 403s after CI, build and provenance all pass.A 403 costs nothing — no artifact uploads, the tag stands, and publish re-runs on the same tag once the entry is fixed. It will be reported immediately for joint escalation rather than retried blindly.
Sweep
git ls-tree origin/master changelog.d/against the range's merges. Every merge carries a fragment except th#1918 itself, which owed one and had none — written in this cut. That matters more than the usual silent-release-note argument: it is the entry a reader will want when they are staring at this release's 403.Verified
assemble_changelog.py --checkclean; 7 fragments consumed,changelog.d/left with onlyREADME.md.uv lock→gen-worker v0.114.2 -> v0.114.3. Range counted bygit rev-list, not estimated.Publish proof dispatched on the branch (
ci.yaml) so its tree matches the merge commit's and the tag does not wait on a fresh post-merge run.