burn-import: add some tests for ConstantNode

[jameshiew]

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

I've been trying to implement OneHot ONNX op (#1714) in a WIP draft branch. The ONNX model ends up containing a constant integer vector values=[0, 1] used by the OneHot op, this vector was causing issues when trying to test the model. I looked at ConstantNode and these are the tests so far I could get working while investigating.

Issues for ConstantNode
#2624 - constant tensors aren't populated with values
#2625 - generated code for const int tensors doesn't compile

Changes

• added a helper method ConstantNode::tensor_ty_tokens for tests, but this PR otherwise shouldn't be changing how ConstantNode currently works
• add codegen tests for ConstantNode (i32/64 + f32/64 scalar, tensors)
• add ONNX model tests for i32/64 + f32/64 scalars - implicitly testing by adding the constant

Testing

Ran added tests

cargo xtask check all
cargo nextest run --manifest-path crates/burn-import/Cargo.toml
cargo nextest run --manifest-path crates/burn-import/onnx-tests/Cargo.toml

I checked the .onnx models contain the expected scalar constants using Netron

Screenshots

[jameshiew]

I'm not sure if the addition is coercing f64 -> f32 somewhere (and i32 -> i64 below). I wasn't sure how to get PyTorch to just forward the constant by itself so these tests are adding the constant to the input

[laggui]

Maybe by having the output return the constant only? But a simple constant addition works too.

In case you're curious, you could also manually define the onnx graph like the ConstOfShape script. PyTorch tends doesn't always have a 1-to-1 correspondence for ops, so in such cases it could be easier to define the graph manually.

I'm not sure if the addition is coercing f64 -> f32 somewhere (and i32 -> i64 below)

The floating point and integer data types are defined by the backend used. A model is not as statically defined like an ONNX graph. If you look at the other tests, the input(s) and output(s) are created using the Tensor methods, not from TensorData.

[codecov]

Codecov Report

Attention: Patch coverage is 99.74160% with 1 line in your changes missing coverage. Please review.

Project coverage is 82.15%. Comparing base (8a89293) to head (756fd3c).
Report is 10 commits behind head on main.

Files with missing lines	Patch %	Lines
crates/burn-import/src/burn/node/constant.rs	99.71%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2623      +/-   ##
==========================================
+ Coverage   82.06%   82.15%   +0.09%     
==========================================
  Files         831      832       +1     
  Lines      106003   106749     +746     
==========================================
+ Hits        86990    87699     +709     
- Misses      19013    19050      +37     

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