Require ImageScaler bias to have one entry per channel

[Akshay Sonawane (apsonawane)]

The bias size check was skipped whenever bias_ was empty, but the compute loop unconditionally reads one bias entry per channel. A model carrying a present-but-empty bias attribute therefore passed validation and then indexed an empty vector. GetAttrs returns OK for a present attribute regardless of element count, so the empty case was reachable from a model file.

An empty bias is not a usable state in either kernel: the constructor already fails outright when the attribute is absent, so requiring the size to equal the channel count keeps the existing contract and closes the gap. The CUDA kernel had the same check and the same per-channel read, so both are updated.

[Copilot]

Pull request overview

This PR tightens runtime validation for the ImageScaler contrib kernels to ensure the bias attribute has exactly one value per input channel, preventing an out-of-bounds read when a model provides a present-but-empty bias list.

Changes:

• Add a regression test that constructs an ImageScaler node with an explicitly empty bias attribute and asserts a validation failure.
• Update the CPU kernel to always require bias_.size() == C (instead of skipping the check when bias_ is empty).
• Update the CUDA kernel with the same always-on size check to match the per-channel bias indexing behavior.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
onnxruntime/test/contrib_ops/tensor_op_test.cc	Adds a test covering the present-but-empty bias attribute case (expects failure).
onnxruntime/contrib_ops/cuda/tensor/image_scaler.cc	Enforces bias length equals channel count in CUDA kernel ComputeInternal.
onnxruntime/contrib_ops/cpu/image_scaler.h	Enforces bias length equals channel count in CPU kernel Compute.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.