Merge pull request #272 from SKaiNET-developers/release/0.6.0

Release/0.6.0

Author: michalharakal

CHANGELOG.md

@@ -1,5 +1,18 @@
 # Changelog
 
+## [0.6.0] - 2025-12-31
+
+### Added
+- StableHLO implementation and E2E CLI app for compiling models to CUDA via IREE.
+- `ArduinoCodegen` for exporting models to standalone C99 code with static memory allocation, optimized for Arduino.
+- KSP-based generation of `TracingOps` for automated recording pipeline updates.
+- Initial implementation of `skainet-compile-hlo` for high-level optimization.
+
+### Changed
+- Improved CUDA backend strategy and IREE integration.
+- Optimized long-running property tests for C code generation.
+- Refactored `TracingTensorOps` to use execution context for code generation.
+
 ## [0.5.1] - 2025-12-26
 
 ### Added

README.md

@@ -99,8 +99,8 @@ Read the [Deep Technical Explanation](docs/arduino-c-codegen.md) for more detail
 
 ```kotlin
 dependencies {
-    implementation("sk.ainet.core:SKaiNET-lang-core:0.5.1")
-    implementation("sk.ainet.core:SKaiNET-backend-cpu:0.5.1")
+    implementation("sk.ainet.core:SKaiNET-lang-core:0.6.0")
+    implementation("sk.ainet.core:SKaiNET-backend-cpu:0.6.0")
 }
 // Ready to build & run in ~8 minutes
 ```
@@ -111,7 +111,7 @@ dependencies {
 - In Kotlin Notebooks for quick exploration
 - With sample projects to learn patterns
 
-See also CHANGELOG for what’s new in 0.5.1.
+See also CHANGELOG for what’s new in 0.6.0.
 
 ## Quick start
 
@@ -126,15 +126,15 @@ dependencyResolutionManagement {
 
 dependencies {
     // minimal dependency with simple CPU backend
-    implementation("sk.ainet.core:SKaiNET-lang-core:0.5.1")
-    implementation("sk.ainet.core:SKaiNET-backend-cpu:0.5.1")
+    implementation("sk.ainet.core:SKaiNET-lang-core:0.6.0")
+    implementation("sk.ainet.core:SKaiNET-backend-cpu:0.6.0")
 
     // simple model zoo
-    implementation("sk.ainet.core:SKaiNET-lang-models:0.5.1")
+    implementation("sk.ainet.core:SKaiNET-lang-models:0.6.0")
 
     // Optional I/O (e.g., GGUF loader, JSON)
-    implementation("sk.ainet.core:SKaiNET-io-core:0.5.1")
-    implementation("sk.ainet.core:SKaiNET-io-gguf:0.5.1")
+    implementation("sk.ainet.core:SKaiNET-io-core:0.6.0")
+    implementation("sk.ainet.core:SKaiNET-io-gguf:0.6.0")
 }
 ```
 
@@ -144,7 +144,7 @@ Maven:
 <dependency>
   <groupId>sk.ainet.core</groupId>
   <artifactId>SKaiNET-lang-core</artifactId>
-  <version>0.5.1</version>
+  <version>0.6.0</version>
 </dependency>
 ```
 
@@ -157,6 +157,34 @@ Maven:
 
 ![Architecture diagram of SKaiNET compiler](docs//SKaiNET-compiler.svg)
 
+## 0.6.0 highlights (with tiny snippets)
+
+- StableHLO and CUDA support via IREE
+
+```kotlin
+// Compile model to StableHLO and run on CUDA
+val ir = Compile.toStableHlo(model)
+println(ir.pretty())
+```
+
+- Arduino C99 code generation
+
+```kotlin
+// Export model to an Arduino library
+val facade = CCodegenFacade()
+facade.exportToArduinoLibrary(
+    model = model,
+    forwardPass = { ctx -> model.forward(input, ctx) },
+    outputPath = "build/arduino",
+    libraryName = "MyModel"
+)
+```
+
+- KSP-based TracingOps generation for recording pipelines.
+- Improved HLO implementation and CUDA backend strategy.
+
+See CHANGELOG.md for the full list.
+
 ## 0.5.0 highlights (with tiny snippets)
 
 - Kolmogorov–Arnold Networks (KAN/AKN) preview in the NN DSL
@@ -213,8 +241,6 @@ println("Tensors: ${gguf.tensors.size}")
 
 - SIMD/Vector API acceleration on JVM; MatMul, tril, pooling ops; forward hooks and simple tape recording; unified tensor creation contexts; nested data blocks returning tensors.
 
-See CHANGELOG.md for the full list.
-
 ## Experimental: Kolmogorov–Arnold Networks (KAN)
 
 SKaiNET includes an initial KAN layer implementation that you can wire into the NN DSL. A KAN layer expands each input feature by a learnable grid of basis coefficients and then mixes them with a linear projection, with optional bias and residual connection.

gradle.properties

@@ -1,5 +1,5 @@
 GROUP=sk.ainet.core
-VERSION_NAME=0.5.1
+VERSION_NAME=0.6.3
 
 POM_DESCRIPTION=SKaiNET
 

skainet-compile/skainet-compile-hlo/gradle.properties

@@ -1,2 +1,2 @@
-POM_ARTIFACT_ID=skainet-compile-dag
+POM_ARTIFACT_ID=skainet-compile-hlo
 POM_NAME=skainet neural network compile DAG

skainet-compile/skainet-compile-onnx/gradle.properties

@@ -1,2 +1,2 @@
-POM_ARTIFACT_ID=skainet-compile-dag
+POM_ARTIFACT_ID=skainet-compile-onnx
 POM_NAME=skainet neural network compile DAG

skainet-lang/skainet-lang-core/build.gradle.kts

@@ -54,10 +54,10 @@ kotlin {
     }
 }
 
-// Ensure KSP metadata task runs before any Kotlin compilation and other KSP tasks
+// Ensure KSP metadata task runs before any Kotlin compilation, other KSP tasks, and sourcesJar
 tasks.configureEach {
     if (name != "kspCommonMainKotlinMetadata" &&
-        (name.startsWith("compileKotlin") || name.startsWith("ksp"))) {
+        (name.startsWith("compileKotlin") || name.startsWith("ksp") || name.contains("ourcesJar"))) {
         dependsOn("kspCommonMainKotlinMetadata")
     }
 }