Skip to content

[RELAX][Frontend][ONNX] Support for SimplifiedLayerNormalization and SkipSimplifiedLayerNormalization contrib ops - #20607

Open
Netra-Srinivasan-McW wants to merge 1 commit into
apache:mainfrom
Netra-Srinivasan-McW:Netra-Srinivasan-McW/onnx-simplified-layer-norm
Open

Netra-Srinivasan-McW wants to merge 1 commit into
apache:mainfrom
Netra-Srinivasan-McW:Netra-Srinivasan-McW/onnx-simplified-layer-norm

Conversation

@Netra-Srinivasan-McW

@Netra-Srinivasan-McW Netra-Srinivasan-McW commented Oct 8, 2026 •

Copy link
Copy Markdown

Summary
This PR adds frontend support for ONNX Runtime contrib operators SimplifiedLayerNormalization and SkipSimplifiedLayerNormalization in TVM Relax. These operators appear in transformer models (such as Xenova/sweep-next-edit-1.5B and SwinliQ-AIs/SmolLM2-1.7B-Instruct) and the absence of these ops raises tvm.error.OpNotImplemented error.

Scripts Involved

  1. python/tvm/relax/frontend/onnx/onnx_frontend.py
  2. tests/python/relax/test_frontend_onnx.py

Testing & Validation
Test modules test_simplified_layer_norm and test_skipsimplifiedlayernormalization validate structural IR equality against expected TVMScript Relax IR for static and dynamic dimensions.

Closes #20625

@Netra-Srinivasan-McW

Netra-Srinivasan-McW commented Oct 9, 2026 •

Copy link
Copy Markdown
Author

@tqchen @tlopex Could you review this PR when you find time. Thanks.

@Netra-Srinivasan-McW

Copy link
Copy Markdown
Author

Hi @mshr-h, @yzhliu, @tqchen, @tlopex,
Could you please take a look at this PR when you get a chance. Thanks.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[Bug][ONNX] SimplifiedLayerNormalization and SkipSimplifiedLayerNormalization ops unsupported

1 participant