ROCm / ROCm/AMDMIGraphX

Add support for 'SplitToSequence' ONNX operator.

Open
#4,286 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

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

Description

DOR (Definition of Ready)

No extra preparations are needed. 'SplitToSequence' is a documented ONNX operator.

Description

FLUX model requires MIGraphX to support 'SplitToSequence' ONNX operator since 'diffusers' version 0.35.0. Probably, it is needed for mapping aten::rms_norm operation into ONNX graph, but I couldn't find confirmation for it in 'diffusers' github.
One needs to support 'SplitToSecond' operator in MIGraphX.

Operator description: https://github.com/onnx/onnx/blob/main/docs/Operators.md#splittosequence

DOD (Definition of Done)

FLUX compilation script works without errors with ONNX opset version 23+ AND with 'diffusers' module of version 0.35.0 or higher. Compiled model produces outputs that are close enough to non-compiled model.

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.

Research direction

Start with the examples/diffusion/python_flux FLUX compilation script and the ONNX Operators.md specification for SplitToSequence. Trace how the operator is handled in MIGraphX and verify the script with ONNX opset 23+ and diffusers 0.35.0 or higher. Done means compilation succeeds and compiled outputs are close to those from the non-compiled model.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp, python
Domain
backend, machine-learning
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.