ROCm / ROCm/AMDMIGraphX

Add SkipGroupNorm Contrib Op

Open
#4,248 0 comments 0 reactions 1 assignee View on GitHub

@TedThemistokleous is already working on this.

Since Aug 20, 2025.

Onnx Operators Perf Improve
Dominant language
C++
Stars
333
Forks
150
Avg merge
4d 19h
Merged PRs (30d)
54

Description

DOR (Definition of Ready)

When we have a change and need to find additional optimizations for stable diffusion pipelines that leverage Onnxruntime as the front end.

Description

Task

Add support for the Microsoft Contrib op SkipGroupNorm
Doc: https://github.com/microsoft/onnxruntime/blob/main/docs/ContribOperators.md#com.microsoft.SkipGroupNorm

Uses

Seen in Stable diffusion optimizations with Onnxruntime's pipeline's

Seen via using the optimize pipeline script for stable diffusion runs

export MIGRAPHX_DISABLE_CK_GEMM=1
export MIGRAPHX_DISABLE_CK_GEMM_FUSION=1
export MIGRAPHX_DISABLE_CK_GEMM_SOFTMAX_GEMM=1

python optimize_pipeline.py  -i ./sd-v2 -o ./sd-v2-0-migx --float16 --disable_attention  --disable_skip_layer_norm   
--disable_embed_layer_norm --disable_bias_skip_layer_norm  --disable_nhwc_conv   --disable_group_norm 

Reference to the workflow on this found here: https://github.com/ROCm/AMDMIGraphX/issues/1688

Using this would give us the ability to further parse OnnxRT optimizations prior to MIGraphX doing any sort of optimizations

DOD (Definition of Done)
  • Operator is parsed in MIGraphX-driver correctly
  • Proper test coverage for operator is given in MIGraphX unit tests
  • Onnxruntime support for operator is present when using MIGraphX Execution Provider

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.