Feat: burn-import implement ONNX ConstantOfShape

[hexd0t]

Pull Request Template

Checklist

• Confirmed that run-checks all script has been executed.
• Made sure the book is up to date with changes in this PR.

Related Issues/PRs

Although ConstantOfShape is not explicitly there, this still benefits #1714

Changes

Add ConstantOfShape support to onnx import.

Testing

Codegen and overall import tests as described in the contributor book.

[codecov]

Codecov Report

Attention: Patch coverage is 89.79592% with 25 lines in your changes missing coverage. Please review.

Project coverage is 85.38%. Comparing base (b331290) to head (739f983).
Report is 9 commits behind head on main.

Files	Patch %	Lines
crates/onnx-ir/src/util.rs	56.00%	11 Missing ⚠️
...tes/burn-import/src/burn/node/constant_of_shape.rs	89.65%	9 Missing ⚠️
crates/burn-import/src/onnx/to_burn.rs	91.17%	3 Missing ⚠️
crates/burn-import/src/burn/ty.rs	93.33%	1 Missing ⚠️
crates/onnx-ir/src/dim_inference.rs	96.29%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1815      +/-   ##
==========================================
- Coverage   85.50%   85.38%   -0.13%     
==========================================
  Files         811      821      +10     
  Lines      100401   101427    +1026     
==========================================
+ Hits        85848    86603     +755     
- Misses      14553    14824     +271     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[laggui]

A couple of changes needed to follow the spec, but otherwise a great start! 💪

[github-actions]

This PR has been marked as stale because it has not been updated for over a month

[antimora]

I will take over this PR and fix issues.

[antimora]

I am trying to resolve some conflicts. It looks like @laggui added ConstantOfShape implementation in "Refactor tensor data" (#1916). I know he had worked on this previously, so it might have been checked in accidentally.

[antimora]

@skewballfox, if you have a chance, please take a look at this PR. I started using Shape arg type instead of using Tensor because shape operation is outside tensor operation but ONNX treats them as if they tensors.

[antimora]

I fixed unresolved issues and started using Shape ArgType which we had but didn't use it. I am approving since I looked at the code and fixed issues.

[laggui]

There's still an issue with the generated ONNX model for the tests.

Otherwise LGTM!

[hexd0t]

Thanks; sorry, I had other things preventing me from finishing this.

[antimora]

Thanks; sorry, I had other things preventing me from finishing this.

No worries! This happens. Thanks for starting it and contributing. Now it's done.