Repository navigation
release: prepare 0.63.0 -- expert heads and routing for Laya, against engine 0.62.0 - #514
Merged
Merged
Conversation
…ardScope pin within four ulp on engine 0.62.0 The scoped-vs-unscoped decode pin was raw-bit identity; since engine 0.62.0 the JVM ops route slab-backed and heap activations through kernels whose accumulation order can differ by one ulp, so the pin allows four ulp.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Release 0.63.0: expert heads and routing for Laya, against engine 0.62.0.
One commit, the release bump: CHANGELOG
[0.63.0] — 2026-10-21(headline over #511 —LayaExpertLoader,LayaDecisionHead,LayaRouterSpec/LayaMixtureAgent,LayaMixtureDecisionAdapter, the ModernBERT-largelaya-typed-decisionscheckpoint; plus the whisper export CLI #510 and the kotest bump #509),VERSION_NAME, the engine pinskainet = "0.62.0"(the JVM ↔ Kotlin/Wasm parity release), the README's current-release paragraph, and the BOM coordinates in the docs (tutorials, reference pages, the Java<version>).Version note carried in the CHANGELOG and README: every API change is additive; the Laya Q8_0 lane on the desktop JVM with
skainet-backend-native-cpuon the classpath needs engine 0.62.1 (SKaiNET #1436 — the native Q8_0 / Q4_0 / Q5_0 / Q5_1 bindings zeroed every row but the last of a multi-row product; found through the LayaMoE example), the FP32 lane and the browser's Q8_0 lane are unaffected.Verified locally against engine 0.62.0 from Maven Central (no local lane):
:llm-inference:laya:jvmTest(12 suites, 45 tests, withLAYA_MULTILINGUAL_PATH,LAYA_TYPED_DECISIONS_PATH,LAYA_MOE_EXPERTS_PATHset:predict.json55 values max 4.99e-5, batched vs single 1.9e-6,LayaQuantizedParityTest45 values max shift 0.0349 5/5 choices,LayaMoeEvalTest9 cases / 306 values max 4.99e-5 with identical routing,sequences.jsonenglish 26/26 and multilingual 26/26, tokenizer corpus 40/40),:decisions:decisions-api:jvmTest,:decisions:decisions-runtime:jvmTest,:llm-inference:laya:compileKotlinWasmJs,jvmApiCheck.The full CI legs run here; tag
0.63.0is pushed after the merge.