[QNN EP] Adjust tolerance for Clip and Transpose tests due to FP16 default in QNN HTP

[yuhuchua-qti]

Description

This PR updates the tolerance thresholds for the Clip and Transpose tests in QnnHTPBackendTests. The adjustment accounts for minor accuracy differences introduced by the change in default floating-point precision in QNN HTP starting from version 2.35.

Motivation and Context

Since QNN 2.35, the default floating-point precision in QNN HTP has changed from FP32 to FP16. Additionally, the configuration option QNN_HTP_GRAPH_CONFIG_OPTION_PRECISION has been deprecated.
This change in precision can lead to expected accuracy loss, especially in scenarios where graph inputs and outputs are defined as float 32, but internal computations are performed in FP16 (e.g., FP32 → FP16 → FP32 conversions). To accommodate this, the tolerance thresholds for the affected tests have been increased to prevent false negatives due to precision differences.

@microsoft-github-policy-service agree company="Qualcomm"

[yuhuchua-qti]

@microsoft-github-policy-service agree company="Qualcomm"

[Copilot]

Pull request overview

This PR adjusts tolerance thresholds for Clip and Transpose operator tests in the QNN HTP backend to accommodate accuracy differences introduced by the default FP16 precision in QNN HTP 2.35+. The precision change affects FP32 models where internal computations now use FP16 (FP32→FP16→FP32 conversions), introducing expected minor accuracy loss.

Key Changes:

• Removes deprecated enable_fp16_precision provider option handling from test helper functions
• Increases tolerance from 1e-5f to 5e-3f for affected float32 tests
• Re-enables previously disabled tests that were failing due to the precision change

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
onnxruntime/test/providers/qnn/transpose_htp_test.cc	Refactored RunTransposeNonQDQOnHTP to accept tolerance parameter; re-enabled and renamed TransposeFloat32OnHTP test with adjusted 5e-3f tolerance
onnxruntime/test/providers/qnn/clip_op_test.cc	Refactored RunClipTest to accept tolerance parameter; re-enabled Clip_f32 test with 5e-3f tolerance; cleaned up GPU test call to use default parameters

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[yuslepukhin]

/azp run Linux QNN CI Pipeline, Win_TRT_Minimal_CUDA_Test_CI, Windows ARM64 QNN CI Pipeline, Windows GPU CUDA CI Pipeline, Windows GPU DML CI Pipeline, Windows GPU Doc Gen CI Pipeline, Windows GPU TensorRT CI Pipeline, Windows OpenVINO CI Pipeline, Windows x64 QNN CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 4 pipeline(s).

[yuslepukhin]

/azp run Linux QNN CI Pipeline, Win_TRT_Minimal_CUDA_Test_CI, Windows ARM64 QNN CI Pipeline, Windows GPU CUDA CI Pipeline, Windows GPU DML CI Pipeline, Windows GPU Doc Gen CI Pipeline, Windows GPU TensorRT CI Pipeline, Windows OpenVINO CI Pipeline, Windows x64 QNN CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 4 pipeline(s).

[edgchen1]

closing and reopening to restart CI pipelines

[yuslepukhin]

Please, address and/or comment on the Copilot review comments so we can get this moving.

[yuhuchua-qti]

Hi @yuslepukhin
I've addressed Copilot's comment.
Thanks.

[yuslepukhin]

/azp run Windows CPU CI Pipeline, Windows GPU CI Pipeline, Windows GPU TensorRT CI Pipeline, ONNX Runtime Web CI Pipeline, Windows ARM64 QNN CI Pipeline, Windows x64 QNN CI Pipeline, onnxruntime-python-checks-ci-pipeline

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[yuslepukhin]

/azp run Linux QNN CI Pipeline, Win_TRT_Minimal_CUDA_Test_CI, Windows ARM64 QNN CI Pipeline, Windows GPU CUDA CI Pipeline, Windows GPU DML CI Pipeline, Windows GPU Doc Gen CI Pipeline, Windows GPU TensorRT CI Pipeline, Windows OpenVINO CI Pipeline, Windows x64 QNN CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 4 pipeline(s).

[yuslepukhin]

/azp run Linux QNN CI Pipeline, Win_TRT_Minimal_CUDA_Test_CI,Windows ARM64 QNN CI Pipeline,Windows GPU Doc Gen CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 4 pipeline(s).

[yuslepukhin]

You are force pushing which makes it impossible to focus on your recent changes. Which in turn makes it hard to find out why CI pipelines are failing. Any particular reason why you do not want to preserve the history at review time?

[yuhuchua-qti]

Hi @yuslepukhin,
Thanks for pointing this out. I didn’t have a particular reason for force‑pushing — it was just out of habit.
I’ll avoid force‑pushing going forward and make sure to preserve history during review.