[ananyaa-ms]

Describe your changes

Add ReplaceTensorsWithConstants, which accepts a map of named ONNX tensors to scalar or array values. It checks each target’s concrete shape and supported numeric or BOOL dtype, infers missing metadata when possible, and validates every replacement before changing the graph. It then uses opset-11-compatible tensor-valued Constant nodes and prunes unused upstream nodes.

Tests cover multi-tensor replacements across dtypes and shapes, shape inference, pruning and runtime output, plus invalid inputs, atomic failure, and opset 11 compatibility.

Checklist before requesting a review

• Add unit tests for this change.
• Make sure all tests can pass.
• Update documents if necessary.
• Lint and apply fixes to your code by running lintrunner -a
• Is this a user-facing change? If yes, give a description of this change to be included in the release notes.

(Optional) Issue link

[Copilot]

🟡 Changes recommended

The surgery can replace unresolved intermediate metadata with an incompatible INT64 tensor and emit an invalid Constant for opset-11 models.

2 open issues

• Validate inferred type and shape before replacing an intermediate · New
• Use an opset-compatible Constant attribute · New

What changed in this PR

Adds ReplaceTensorWithConstant to Olive’s ONNX graph surgeries, allowing a named scalar INT64 tensor’s producer to be replaced with a constant while pruning unused upstream nodes.

Changes:

• Adds tensor lookup, producer validation, and signed 64-bit value validation.
• Preserves shared outputs while removing unused computation.
• Adds runtime checks and tests for invalid targets and replacement values.

File	Description
test/​passes/​onnx/​test_graph_surgeries.py	Tests pruning, shared-output preservation, runtime results, and invalid inputs.
olive/​passes/​onnx/​graph_surgeries.py	Implements constant replacement and unused-node cleanup.

_{Review effort: Balanced}

[Justin Chu (justinchuby)]

I feel like this is something that can be generalized. For example, would it make sense if you allow a map as input, so users don't need to run the pass multiple times to replace tensors? Additionally, it seems like we can just allow users to replace some value with some arbitrary dtype matching constant?

[ananyaa-ms]

I feel like this is something that can be generalized. For example, would it make sense if you allow a map as input, so users don't need to run the pass multiple times to replace tensors? Additionally, it seems like we can just allow users to replace some value with some arbitrary dtype matching constant?

Updated the pass to use a map to replace multiple tensors, and to accept any dtype the installed ONNX libraries can safely encode from JSON