Repository navigation
Prepare Models 0.3.60 - #268
Merged
Merged
Conversation
Ships the router changes merged in #267: a public `ModelCandidate.costOf` the shared budget delegates to, `PretrainedTaskClassifier.match`/`admits` so a caller can see how confident a classification was, an optional per-candidate `maxOutputTokens` because one execution-wide bound cannot describe a fleet holding both a liveness model and a reasoning model, and a token-bound refusal that carries the completion it was billed for instead of discarding it. `verifyReleaseMetadata` checks twenty-odd places the version appears and found nine of them: the Antora component and its models-version attribute, docs package.json and package-lock.json, the landing page pill, the native crate's Cargo.toml and Cargo.lock, the notebook .env.example, docker-compose, prepare-classpath.sh and README, and published coordinates in the models-backend-apple and models-rag READMEs. All are at 0.3.60. The 0.3.59 changelog heading stays where it is. Consumes released Vectors 0.1.30, unchanged. No CUDA numeric path changed and no device gate was rerun, which the changelog states along with the last G1 and G4 results. No model qualification changed. Verified: test, build, spotlessCheck, verifyReleaseMetadata, verifyDependencyPolicy, complianceCheck and jacocoTestCoverageVerification.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Release preparation for 0.3.60, shipping the router changes merged in #267.
ModelCandidate.costOf, withRoutingBudgetdelegating to itPretrainedTaskClassifier.match/admits, so a weak match is distinguishable from no matchmaxOutputTokens, since one execution-wide bound cannot describe a fleet holding both a liveness model and a reasoning modelbilledResult(),billedUsage()andbilledCost()rather than discarding a completion the provider already charged forverifyReleaseMetadatafound nine stale version pins beyond gradle.properties; all are updated. Consumes released Vectors 0.1.30. No CUDA numeric path changed and no device gate was rerun; the changelog records the last G1 and G4 results and that no model qualification changed.Verified locally:
test,build,spotlessCheck,verifyReleaseMetadata,verifyDependencyPolicy,complianceCheck,jacocoTestCoverageVerification.