feat(delphi): Q10 carve-out — pin large-conv cutoffs to certify the deterministic full-PCA path#2652
Draft
jucor wants to merge 1 commit into
Draft
feat(delphi): Q10 carve-out — pin large-conv cutoffs to certify the deterministic full-PCA path#2652jucor wants to merge 1 commit into
jucor wants to merge 1 commit into
Conversation
This was referenced Jul 22, 2026
Draft
docs(delphi): R1-parity goal docs, journal (2026-07-18 → 07-24 s5), quirks + divergence ledger
#2626
Draft
Draft
…eterministic full-PCA path
Clojure's conv-update switches to mini-batch partial-pca on an UNSEEDED
Mersenne-Twister row sample when n-ptpts > 10000 or n-cmts > 5000
(conversation.clj:757-815) — no deterministic reference exists on that
path, even between two Clojure runs (CLOJURE_QUIRKS.md Q10; root cause of
the pakistan/engage/bg2050 mid-chain battery divergences, verified to
switch at exactly the first divergent step on all three).
The replay driver now passes {:ptpt-cutoff :cmt-cutoff} = 10^9 into
conv-update so certification exercises the full-PCA path at every size
(math/dev/ change only — no math sources touched), and certify.py's
acceptance notice logs the carve-out on every run, like Q7's.
Python keeps full PCA at all sizes (no change needed).
commit-id:7c20f8e1
Contributor
There was a problem hiding this comment.
Pull request overview
This PR makes the Clojure replay/certification reference deterministic for very large conversations by forcing conv-update to stay on the full-PCA (“small-conv”) path regardless of participant/comment counts, avoiding the unseeded mini-batch PCA sampling path (CLOJURE_QUIRKS Q10). It also updates the certification acceptance notice so every run explicitly documents that this carve-out is in effect.
Changes:
- Pin
:ptpt-cutoff/:cmt-cutoffto10^9in the Clojure Mode A replay driver and pass these options intoconv/conv-updateon every step. - Emit a per-run stderr log line in the Clojure driver indicating the Q10 carve-out is active.
- Expand
certify.py’sACCEPTANCE_NOTICEto include the Q10 carve-out statement alongside the existing Q7 notice.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.
| File | Description |
|---|---|
| math/dev/replay.clj | Forces deterministic full-PCA path during Clojure replay by passing very large cutoffs into conv-update, and logs the Q10 carve-out. |
| delphi/polismath/replay/certify.py | Updates the always-printed acceptance notice to document the Q10 carve-out during certification runs. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
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.
Clojure's conv-update switches to mini-batch partial-pca on an UNSEEDED
Mersenne-Twister row sample when n-ptpts > 10000 or n-cmts > 5000
(conversation.clj:757-815) — no deterministic reference exists on that
path, even between two Clojure runs (CLOJURE_QUIRKS.md Q10; root cause of
the pakistan/engage/bg2050 mid-chain battery divergences, verified to
switch at exactly the first divergent step on all three).
The replay driver now passes {:ptpt-cutoff :cmt-cutoff} = 10^9 into
conv-update so certification exercises the full-PCA path at every size
(math/dev/ change only — no math sources touched), and certify.py's
acceptance notice logs the carve-out on every run, like Q7's.
Python keeps full PCA at all sizes (no change needed).
commit-id:7c20f8e1
Stack: