release: prepare 0.61.0 -- SKEEP-007 browser lane

CHANGELOG.md

@@ -2,7 +2,25 @@
 
 ## [Unreleased]
 
-Headline: **the browser lane stops falling to the reference kernel (SKEEP-007, Tier 0).**
+## [0.61.0] - 2026-10-16
+
+Headline: **the browser lane (SKEEP-007): packed SIMD128 kernels, quantize-at-load, and no more
+reference-kernel fallbacks on wasm.** Every SKaiNET browser build used to run scalar Kotlin on one
+thread and sent canonical GGUF weights to the per-element decoding reference kernel. This release
+reorders the packed-weight paths so the scalar packed kernels serve every target (Tier 0), ships the
+engine's own C kernels as a SIMD128 sidecar for wasmJs with pthreads on cross-origin-isolated pages
+(Tier 1, `skainet-backend-wasm-simd`), and teaches the SafeTensors and GGUF loaders to quantize dense
+weights at load (Tier 2), so Hugging Face checkpoints can be served packed on every target. Measured
+on Node: Laya's Q8_0 matmuls move from 0.72 GFLOPS to 27 GFLOPS with four threads, one 22-layer
+decision from about 29 s to about 1.2 s.
+
+Version note: the SKEEP-006 compatibility aliases scheduled for removal in 0.61 are removed (see
+*Removed*). SKaiNET-transformers follows with the Laya consumer of quantize-at-load and the browser
+pack.
+
+### Removed
+
+- **The SKEEP-006 compatibility surface**, an error since 0.60 as scheduled: the `sk.ainet.io.tokenizer.*` aliases (`Tokenizer`, `QwenByteLevelBpeTokenizer`, `SentencePieceTokenizer`, `SpecialTokenSplitter`, `TekkenTokenizer`, `ByteToUnicode`), `sk.ainet.io.model.ModelFormat`, `sk.ainet.io.weights.NameMap` / `RoleTableNameMap` / `TransformerNameMaps`, and in `skainet-io-gguf` the `sk.ainet.io.gguf.registry` package (`ModelArchitecture.fromGguf`, `ModelLoaderRegistry`, `ModelLoaderDescriptor`, `LlamaLoaderDescriptor`, `Gemma3nLoaderDescriptor`). The homes are `sk.ainet.frontend.text`, `sk.ainet.frontend.tokenizer` and `sk.ainet.model`; `TokenizerFactory` and `UnsupportedTokenizerException` stay in `sk.ainet.io.tokenizer` for Java callers, and `sk.ainet.io.weights.asWeightNameResolver` stays.
 
 ### Added
 

CITATION.cff

@@ -12,8 +12,8 @@ authors:
     orcid: https://orcid.org/0009-0006-7981-4075
   - name: SKaiNET-developers contributors
     website: https://github.com/SKaiNET-developers
-version: "0.60.0"
-date-released: "2026-10-15"
+version: "0.61.0"
+date-released: "2026-10-16"
 license: MIT
 repository-code: https://github.com/SKaiNET-developers/SKaiNET
 url: https://skainet-developers.github.io/SKaiNET/

README.md

@@ -51,7 +51,7 @@ Add the core dependencies (Gradle Kotlin DSL):
 ```kotlin
 dependencies {
     // Recommended: import the umbrella BOM and drop versions on the engine modules.
-    implementation(platform("sk.ainet:skainet-bom:0.60.0"))
+    implementation(platform("sk.ainet:skainet-bom:0.61.0"))
 
     implementation("sk.ainet.core:skainet-lang-core")
     implementation("sk.ainet.core:skainet-backend-cpu")
@@ -308,27 +308,26 @@ val withoutLabel = dataPipeline<RawDataset>()
 
 ---
 
-## What's New in 0.60.0
-
-Key-padding lengths and window metadata for attention (#1343):
-
-- **`scaledDotProductAttention(keyLengths, window, rightContext)`** — padding (a `[batch]`
-  lengths tensor), the sliding-window band and `causal` are rules over positions defined once in
-  `AttentionMask` and shared by the CPU kernel, the recorded graph and the StableHLO export. The
-  CPU kernel computes scores only inside each row's attended key range, so a padded batch equals
-  the unpadded single passes; the dense additive mask stays the generic fallback.
-- **The rules travel as op metadata** — the recorded node carries `maskKind`, `window` and
-  `rightContext` with the lengths as an operand, so a compiled target regenerates the cheapest
-  legal form (`iota`/`compare`/`select` on the export) instead of inheriting a dense
-  `[batch, 1, Sq, Sk]` mask that scales with the batch.
-- **`skainet.sdpa` composite** — the composite seam is n-ary (`HloDecomposition` takes every
-  operand), and `SdpaLowering.spec` emits `stablehlo.composite "skainet.sdpa"` with `scale`,
-  `causal`, `window`, `rightContext` and `maskKind` attributes when a target's
-  `OpGranularityPolicy` keeps attention fused, from the same builder as the inline form.
-- The SKEEP-006 compatibility aliases (`sk.ainet.io.tokenizer.*`, `sk.ainet.io.model.ModelFormat`,
-  `sk.ainet.io.weights.NameMap`, the GGUF `ModelLoaderRegistry`) are errors since this release and
-  are removed in 0.61. Fixed on the way: a batched dense attention mask on the CPU read batch
-  row 0 for every batch.
+## What's New in 0.61.0
+
+The browser lane (SKEEP-007): packed SIMD128 kernels, quantize-at-load, no reference-kernel
+fallbacks on wasm.
+
+- **`skainet-backend-wasm-simd`** — the engine's own C kernels compiled by Emscripten into a
+  SIMD128 sidecar for wasmJs, with pthreads on cross-origin-isolated pages and under Node. One call,
+  `WasmSimd.install()`, registers the kernel pack and the provider and makes every context built
+  afterwards use them. Measured on Node: Laya's Q8_0 matmuls from 0.72 GFLOPS (scalar) to 8.5
+  single-threaded and 27 with four threads — a 22-layer decision from about 29 s to about 1.2 s.
+- **Quantize at load** — `WeightForm(RequantizeTo(Q8_0))` on the SafeTensors loaders (and every
+  encoding with a quantizer on the GGUF loader) turns dense F32/F16/BF16 weight matrices into packed
+  Q8_0 or Q4_0 at load; `PlannerProfile.quantizeDenseTo` lets a browser or phone profile ask for it;
+  `skainet-quantize` converts offline. Hugging Face checkpoints can be served packed on every target.
+- **The scalar lane is no longer the reference lane** — a canonical packed weight on a target without
+  a row-major kernel pack is relaid once into kernel feed order and served by the scalar packed
+  kernels; dense `[out, in]` weights are multiplied in place without a transpose copy; 16-bit
+  token tables are read row by row. `ByteArrayRandomAccessSource` is the engine's in-memory source.
+- The SKEEP-006 compatibility aliases, errors since 0.60, are removed as scheduled; `TokenizerFactory`
+  stays in `sk.ainet.io.tokenizer` for Java callers.
 
 See [CHANGELOG.md](CHANGELOG.md) for full release notes, including every prior release.
 
@@ -354,6 +353,13 @@ We love contributions! Whether it's a new operator, documentation, or a bug fix:
 
 Browse the full codebase documentation on [DeepWiki](https://deepwiki.com/SKaiNET-developers/SKaiNET).
 
+### Contributors (0.61.0)
+
+- **Michal Harakal** ([@michalharakal](https://github.com/michalharakal)) — SKEEP-007: the Tier 0
+  dispatch-order and dense-path fixes, quantize-at-load for the SafeTensors and GGUF loaders with the
+  Q8_0 quantizer and `skainet-quantize`, the `skainet-backend-wasm-simd` sidecar kernel pack, the
+  scheduled SKEEP-006 removal, and the release
+
 ### Contributors (0.60.0)
 
 - **Michal Harakal** ([@michalharakal](https://github.com/michalharakal)) — key-padding lengths
@@ -456,7 +462,7 @@ from it). Ready-to-paste BibTeX:
   author       = {Harakal, Michal and {SKaiNET-developers contributors}},
   title        = {{SKaiNET}: A {Kotlin} Multiplatform {AI} Framework},
   year         = {2026},
-  note         = {Version 0.60.0. MIT License},
+  note         = {Version 0.61.0. MIT License},
   howpublished = {\url{https://github.com/SKaiNET-developers/SKaiNET}}
 }
 ```

docs/antora.yml

@@ -15,7 +15,7 @@ asciidoc:
     framework_name: SKaiNET
     # Current SKaiNET release — bump once per release; referenced as
     # {skainet_version} in dependency snippets (blocks need subs="attributes+").
-    skainet_version: 0.60.0
+    skainet_version: 0.61.0
     ksp_version: 2.2.21-2.0.5
     dokka_version: 2.1.0
     asciidoctorj_version: 3.0.0

docs/modules/ROOT/pages/reference/input-frontends.adoc

@@ -30,8 +30,8 @@ own side files and decorated by its model family.
 |===
 
 `skainet-io-core` keeps `sk.ainet.io.tokenizer.*`, `sk.ainet.io.model.ModelFormat` and
-`sk.ainet.io.weights.NameMap` as deprecated aliases (WARNING in 0.58 and 0.59, ERROR since 0.60,
-removed in 0.61), and `TokenizerFactory` as a forwarding object for Java callers.
+`sk.ainet.io.weights.NameMap` as deprecated aliases for two releases (WARNING in 0.58 and 0.59, ERROR
+in 0.60, removed in 0.61), and keeps `TokenizerFactory` as a forwarding object for Java callers.
 
 == Resolving a checkpoint's tokenizer
 

docs/modules/ROOT/pages/reference/kernel-support-matrix.adoc

@@ -1,7 +1,7 @@
 = Kernel × platform support matrix
 :description: Which compute-kernel provider serves each weight format on each KMP target.
 
-Generated from `kernel-support.json` (version `0.60.0`) by `KernelSupportMatrixTest` — registry introspection of the registered `KernelProvider` implementations. Do not edit by hand; run `./gradlew generateKernelMatrix` to refresh.
+Generated from `kernel-support.json` (version `0.61.0`) by `KernelSupportMatrixTest` — registry introspection of the registered `KernelProvider` implementations. Do not edit by hand; run `./gradlew generateKernelMatrix` to refresh.
 
 Each cell is the best (highest-priority) provider that serves `Float32 × format` `matmul` on that platform: *native-ffm* (100) → *panama-vector* (50) → *scalar* (0). An empty cell (`—`) means no provider carries a kernel there (the format is dequant-to-FP32 only).
 

docs/modules/ROOT/pages/reference/operators/generated/index.adoc

@@ -1,6 +1,6 @@
 = AI-NET Operators Reference
 
-Generated from version `0.60.0` on 2026-10-15
+Generated from version `0.61.0` on 2026-10-16
 
 == Operators by Modality
 

docs/modules/ROOT/pages/reference/ops-status-matrix.adoc

@@ -1,7 +1,7 @@
 = Operator Coverage Matrix
 :description: Cross-backend status for every operator function in SKaiNET.
 
-Generated from `operators.json` version `0.60.0` on 2026-10-15.
+Generated from `operators.json` version `0.61.0` on 2026-10-16.
 
 Rows are `Operator.function` pairs. The `Validated` column shows whether the function's documentation has been DARC-validated by a reviewer (see xref:contributing/darc-workflow.adoc[DARC workflow]). The `Lowering` column lists the compile-path forms the function declares with `@Lowering` (`composite` = a `stablehlo.composite` when a target keeps it fused, `decompose` = StableHLO/CHLO primitives, `native` = a 1:1 target op); `—` means undeclared. Remaining columns are backends that appear in any function's `statusByBackend` map — a missing entry means the backend makes no claim about the function (treat it as "unknown", not "not supported").
 

docs/modules/ROOT/pages/tutorials/android-classifier-getting-started.adoc

@@ -19,12 +19,12 @@ An Android project with Kotlin. Add the SKaiNET modules:
 [source,kotlin]
 ----
 dependencies {
-    implementation("sk.ainet.core:skainet-lang-core:0.60.0")     // tensors, DSL, training
-    implementation("sk.ainet.core:skainet-backend-cpu:0.60.0")   // CPU ops
-    implementation("sk.ainet.core:skainet-compile-dag:0.60.0")   // autograd (training context)
-    implementation("sk.ainet.core:skainet-data-api:0.60.0")      // Dataset / DataBatch
-    implementation("sk.ainet.core:skainet-data-simple:0.60.0")   // embedded Iris
-    runtimeOnly("sk.ainet.core:skainet-backend-jni-cpu:0.60.0")  // NEON kernels (see below)
+    implementation("sk.ainet.core:skainet-lang-core:0.61.0")     // tensors, DSL, training
+    implementation("sk.ainet.core:skainet-backend-cpu:0.61.0")   // CPU ops
+    implementation("sk.ainet.core:skainet-compile-dag:0.61.0")   // autograd (training context)
+    implementation("sk.ainet.core:skainet-data-api:0.61.0")      // Dataset / DataBatch
+    implementation("sk.ainet.core:skainet-data-simple:0.61.0")   // embedded Iris
+    runtimeOnly("sk.ainet.core:skainet-backend-jni-cpu:0.61.0")  // NEON kernels (see below)
 }
 ----
 

docs/modules/skeep/pages/006-input-frontends-and-model-identity.adoc

@@ -2,7 +2,7 @@
 :description: SKaiNET proposal for one resolvable input front-end contract (tokenizers first, audio and image features next) keyed on a string-based model identity with an immutable family registry, so that every model type SKaiNET runs gets its preprocessing from the engine instead of re-deriving it downstream.
 
 Status: Implemented +
-Implementation: https://github.com/SKaiNET-developers/SKaiNET/pull/1332[#1332] (text lane, identity, resolver; engine 0.58.0), https://github.com/SKaiNET-developers/SKaiNET-transformers/pull/485[SKaiNET-transformers #485] (consumer, 0.58.0); the io-core compatibility aliases moved to `DeprecationLevel.ERROR` in 0.60.0 (https://github.com/SKaiNET-developers/SKaiNET/pull/1350[#1350]) and leave in 0.61. The audio (F4, F6) and image (F5) front ends are follow-up proposals. +
+Implementation: https://github.com/SKaiNET-developers/SKaiNET/pull/1332[#1332] (text lane, identity, resolver; engine 0.58.0), https://github.com/SKaiNET-developers/SKaiNET-transformers/pull/485[SKaiNET-transformers #485] (consumer, 0.58.0); the io-core compatibility aliases moved to `DeprecationLevel.ERROR` in 0.60.0 (https://github.com/SKaiNET-developers/SKaiNET/pull/1350[#1350]) and were removed in 0.61.0. The audio (F4, F6) and image (F5) front ends are follow-up proposals. +
 Audience: SKaiNET maintainers and contributors; SKaiNET-transformers maintainers (the first consumer, and the repository whose parallel tokenizer stack this proposal retires); SKaiNET-cartridge maintainers (the facade binds front ends to cartridge ports) +
 Created: 2026-10-11 +
 Tracking issue: https://github.com/SKaiNET-developers/SKaiNET/issues/1331[#1331] (implementation lanes: https://github.com/SKaiNET-developers/SKaiNET/issues/1321[#1321], https://github.com/SKaiNET-developers/SKaiNET-transformers/issues/471[SKaiNET-transformers #471], https://github.com/SKaiNET-developers/SKaiNET-transformers/issues/330[SKaiNET-transformers #330]) +

docs/modules/skeep/pages/007-browser-lane.adoc

@@ -1,8 +1,8 @@
 = SKEEP-007: The browser lane — packed SIMD kernels and quantize-at-load
 :description: SKaiNET proposal for making inference in a browser usable and quick: stop the scalar wasm lane from falling to the decoding reference kernel, serve dense weights without transpose copies, keep token tables narrow, add a SIMD128 kernel pack with pthreads as a sidecar wasm module, and let the SafeTensors loader quantize at load so Hugging Face checkpoints are served packed on every target.
 
-Status: Draft — Tiers 0, 2 and 1 in review +
-Implementation: Tier 0 https://github.com/SKaiNET-developers/SKaiNET/pull/1352[#1352]; Tier 2 quantize-at-load https://github.com/SKaiNET-developers/SKaiNET/pull/1353[#1353]; Tier 1 `skainet-backend-wasm-simd` (the pull request stacked on #1353); the consumer work in SKaiNET-transformers follows, see <<_rollout_plan>>. +
+Status: Implemented (Tiers 0, 1 and 2 in engine 0.61.0); the resident-activation step, the SKaiNET-transformers consumer and the WebGPU proposal remain open +
+Implementation: Tier 0 https://github.com/SKaiNET-developers/SKaiNET/pull/1352[#1352], Tier 2 quantize-at-load https://github.com/SKaiNET-developers/SKaiNET/pull/1353[#1353], Tier 1 `skainet-backend-wasm-simd` https://github.com/SKaiNET-developers/SKaiNET/pull/1354[#1354] — engine 0.61.0; the consumer work in SKaiNET-transformers follows, see <<_rollout_plan>>. +
 Audience: SKaiNET maintainers and contributors; SKaiNET-transformers maintainers (Laya and BERT in the browser); SKaiNET-examples maintainers (every web sample runs through this lane) +
 Created: 2026-10-16 +
 Origin: the SemanticEmojiFilter example ships its 23M-parameter embedding filter to the web but keeps the 322M-parameter Laya verifier on desktop and Android, because a decision in scalar Kotlin/Wasm is estimated at 10–30 s and the FP32 weights at about 2.5 GB resident. Nothing in the repository measures the wasm lane today.

docs/modules/skeep/pages/index.adoc

@@ -136,6 +136,6 @@ Every proposal should include:
 | Input front ends and model identity — one resolvable preprocessing contract (tokenizers first) keyed on a string-based family registry
 
 | xref:skeep:007-browser-lane.adoc[SKEEP-007]
-| Draft
+| Implemented
 | The browser lane — packed SIMD128 kernels as a sidecar module, quantize-at-load for SafeTensors, and the common-code fixes that keep wasm off the reference kernel
 |===

gradle.properties

@@ -1,5 +1,5 @@
 GROUP=sk.ainet.core
-VERSION_NAME=0.60.0
+VERSION_NAME=0.61.0
 POM_DESCRIPTION=SKaiNET
 
 POM_URL=https://github.com/SKaiNET-developers/skainet/

skainet-io/skainet-io-core/src/commonMain/kotlin/sk/ainet/io/weights/NameMap.kt

@@ -1,17 +1,10 @@
-@file:Suppress("DEPRECATION", "DEPRECATION_ERROR")
 
 package sk.ainet.io.weights
 
 import sk.ainet.lang.tensor.TensorId
 
 /** Moved to `skainet-model-identity` (SKEEP-006); alias an error since 0.60, removed in 0.61. */
-@Deprecated("Moved to sk.ainet.model.NameMap (SKEEP-006); the alias is an error since 0.60 and is removed in 0.61.", ReplaceWith("NameMap", "sk.ainet.model.NameMap"), level = DeprecationLevel.ERROR)
-public typealias NameMap = sk.ainet.model.NameMap
-
 /** Moved to `skainet-model-identity` (SKEEP-006); alias an error since 0.60, removed in 0.61. */
-@Deprecated("Moved to sk.ainet.model.RoleTableNameMap (SKEEP-006); the alias is an error since 0.60 and is removed in 0.61.", ReplaceWith("RoleTableNameMap", "sk.ainet.model.RoleTableNameMap"), level = DeprecationLevel.ERROR)
-public typealias RoleTableNameMap = sk.ainet.model.RoleTableNameMap
-
 /**
  * This map as a legacy [WeightNameResolver]: `(modulePath, paramName)` in the slash form
  * (`model/layers[3]/attn`, `q_proj.weight`) → checkpoint name. An extension since SKEEP-006