Skip to content

Optimizer Suggestions for Common Graph Fusions #2301

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
AyoubMDL opened this issue May 13, 2025 · 1 comment
Open

Optimizer Suggestions for Common Graph Fusions #2301

AyoubMDL opened this issue May 13, 2025 · 1 comment
Assignees
@AyoubMDL
Labels
contribution welcome We welcome code contributions for this good first issue Good for newcomers module: rewriter

Comments

@AyoubMDL
Copy link
Contributor

AyoubMDL commented May 13, 2025 •
edited
Loading

I've been using ONNXScript extensively lately and really appreciate how powerful and flexible it is. As part of that experience, I and @Johansmm would like to suggest a few optimization patterns that could improve performance and simplify common model graphs. These are generally applicable across many models:

  1. Fold BatchNormalization into preceding nodes (Conv, ConvTranspose, Gemm) Rewriter: Fold Batchnorm nodes #2312
  2. Fuse MatMul + Add into Gemm
  3. Fuse Min + Max into Clip
  4. Eliminate redundant Reshape / Flatten nodes
  5. Fold Pad into Conv / ConvTranspose
  6. Fuse successive Clip / ReLU nodes

We're happy to contribute if any of these are a good fit. Feel free to add other suggestions or let us know if some of these are out of scope.
@justinchuby justinchuby added module: rewriter good first issue Good for newcomers contribution welcome We welcome code contributions for this labels May 13, 2025
@justinchuby
Copy link
Collaborator

Thanks! All of the above sounds good. PRs welcomed. For (2) we want it to be optional because some runtimes prefer one way over another. You may add the rules in https://github.com/microsoft/onnxscript/tree/main/onnxscript/rewriter and use the existing ones as reference.

@AyoubMDL AyoubMDL mentioned this issue May 18, 2025
Merged
Johansmm added a commit to Johansmm/onnxscript that referenced this issue May 22, 2025
@Johansmm
[Rewriter]: introduce fuse_pad_into_conv (microsoft#2301)
ae99c8d
Johansmm added a commit to Johansmm/onnxscript that referenced this issue May 26, 2025
@Johansmm
[Rewriter]: introduce fuse_pad_into_conv (microsoft#2301)
99dfbef
Johansmm added a commit to Johansmm/onnxscript that referenced this issue May 26, 2025
@Johansmm
[Rewriter]: introduce fuse_pad_into_conv (microsoft#2301)
6e4b326
justinchuby pushed a commit that referenced this issue May 27, 2025
@AyoubMDL
Rewriter: Fold Batchnorm nodes (#2312)
276bf27 
Fuses `BatchNormalization` nodes into the following nodes (`Conv`,
`ConvTranspose`, `Gemm`)
(#2301)
Johansmm added a commit to Johansmm/onnxscript that referenced this issue May 30, 2025
@Johansmm
[Rewriter]: introduce fuse_pad_into_conv (microsoft#2301)
a8c5f52
Johansmm added a commit to Johansmm/onnxscript that referenced this issue May 30, 2025
@Johansmm
[Rewriter]: introduce fuse_pad_into_conv_transpose (microsoft#2301)
1f956be
Johansmm added a commit to Johansmm/onnxscript that referenced this issue May 30, 2025
@Johansmm
[Rewriter]: introduce fuse_pad_into_conv_integer (microsoft#2301)
3be311c
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@AyoubMDL AyoubMDL

Labels
contribution welcome We welcome code contributions for this good first issue Good for newcomers module: rewriter
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@justinchuby @AyoubMDL