feat(speculation): wire predictor into speculation pipeline - #684
Open
behinddwalls wants to merge 5 commits into
Open
feat(speculation): wire predictor into speculation pipeline#684behinddwalls wants to merge 5 commits into
behinddwalls wants to merge 5 commits into
Conversation
This was referenced Sep 7, 2026
behinddwalls
force-pushed
the
preetam/outcome-predictor-wiring
branch
from
September 8, 2026 00:13
caeb62b to
3c5ee44
Compare
behinddwalls
force-pushed
the
preetam/outcome-predictor-wiring
branch
2 times, most recently
from
September 8, 2026 00:51
61d1369 to
33212d5
Compare
## Summary ### Why? The predictor implementation is additive until the speculation pipeline supplies each dependency's run-local path evidence and uses the revised probability for ranking. ### What? Thread path sets through the Generator contract and standard Speculator, replace `bestfirst`'s scorer dependency with the predictor, and compose the evidence predictor from per-queue YAML configuration in orchestrator profiles. Neutral default factors preserve scorer-only ranking when no factors are configured. ## Test Plan - `bazel test //submitqueue/extension/speculation/generator/... //submitqueue/extension/speculation/speculator/... //service/submitqueue/orchestrator/server:go_default_test` - `make check-gazelle`
## Summary ### Why? The orchestrator configuration comments still described factors as odds multipliers and referenced fitting work removed from the RFC. ### What? Describe factors directly as revisions to the scorer price and retain the RFC's neutral, positive-factor contract without changing configuration behavior. ## Test Plan - ✅ `./tool/bazel test //submitqueue/extension/speculation/predictor/... //submitqueue/extension/speculation/generator/... //submitqueue/extension/speculation/speculator/... //service/submitqueue/orchestrator/server:go_default_test` - ✅ `make check-gazelle`
## Summary ### Why? Configuration accepted infinite evidence factors even though the predictor rejects them, and the speculator composition test did not prove that path-set evidence reaches the Generator. ### What? Reject non-finite factors during profile loading, cover infinite YAML values, and require the exact path-set snapshot in the Generator wiring expectation. ## Test Plan - ✅ `make fmt` - ✅ `./tool/bazel test //submitqueue/extension/speculation/predictor/... //submitqueue/extension/speculation/generator/... //submitqueue/extension/speculation/speculator/... //service/submitqueue/orchestrator/server:go_default_test`
A queue predictor block now revises named factors instead of replacing the whole map, so defaults like pathFailed stay in force. Best-first tests rank pathFailed, cancelling, and merging through the evidence predictor rather than a stub.
Update generator and standard-speculator guides for the shared path-set snapshot and predictor-backed best-first ranking.
behinddwalls
force-pushed
the
preetam/outcome-predictor-wiring
branch
from
September 8, 2026 00:57
33212d5 to
a97225b
Compare
behinddwalls
marked this pull request as ready for review
September 8, 2026 01:09
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.
Summary
Why?
The predictor implementation is additive until the speculation pipeline supplies each dependency's run-local path evidence and uses the revised probability for ranking. A queue
predictorblock also used to replace the whole factors map, so naming one factor silently reset the rest of the defaults to1.What?
Thread path sets through the Generator contract and standard Speculator, replace
bestfirst's scorer dependency with the predictor, and compose the evidence predictor from per-queue YAML configuration in orchestrator profiles. Queue predictor blocks overlay named factors onto defaults, profile loading rejects non-finite factors, and best-first ranking tests coverpathPassed,pathFailed,cancelling, andmergingthrough the evidence predictor. Update the generator, best-first, and standard-speculator guides alongside the implementation they describe.Test Plan
./tool/bazel test //submitqueue/extension/speculation/generator/bestfirst:go_default_test //service/submitqueue/orchestrator/server:go_default_testmake fmtIssues
Stack