Register silu / SiLU alias in ActivationOperationsConverter (#484)#486
Merged
michalharakal merged 1 commit intodevelopfrom Apr 13, 2026
Merged
Register silu / SiLU alias in ActivationOperationsConverter (#484)#486michalharakal merged 1 commit intodevelopfrom
michalharakal merged 1 commit intodevelopfrom
Conversation
…484) Every Llama / Mistral / Qwen / Gemma family model names the same x * sigmoid(x) activation `silu` rather than `swish`. PyTorch's F.silu and HuggingFace's ACT2FN["silu"] both emit the `silu` name, so a traced LLM forward pass was falling through the StableHLO converter registry's "no converter found" path. Fix is scoped: - Adds `silu` and `SiLU` to supportedOperations. - Routes `silu` to the existing convertSwish lowering in the dispatch `when`. Same op bodies, identical MLIR output. - Extends ActivationOperationsConverterTest's registration check with `silu` / `SiLU` and adds testSiluAliasLowersLikeSwish which asserts the emitted module carries the same ops (negate, exp, const 1.0, add, divide, multiply) that the swish test already asserts. Verified locally with `./gradlew :skainet-compile:skainet-compile-hlo:allTests -x kotlinWasmStoreYarnLock` — green across jvmTest, wasmJsTest, wasmJsBrowserTest, wasmWasiTest, wasmWasiNodeTest, macosArm64Test, and iosSimulatorArm64Test. Running allTests (not just jvmTest) is now standing discipline for this repo after #485 caught a JVM-only stdlib call that broke KMP compile. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes #484.
Summary
Every Llama / Mistral / Qwen / Gemma family model names the same `x * sigmoid(x)` activation `silu` (Sigmoid Linear Unit), not `swish`. PyTorch's `F.silu` and HuggingFace's `ACT2FN["silu"]` both emit the `silu` name. Without this alias, a traced Llama forward pass falls through the StableHLO converter registry's "no converter found" path.
Change
Test plan
🤖 Generated with Claude Code