QNN Execution Provider Backend Path Warning

[leestott]

QNN Execution Provider Backend Path Warning

Severity: Warning (non-blocking)
Component: ONNX Runtime QNN Execution Provider
Reproduction: Loading any model on Snapdragon hardware with NPU support

[W:onnxruntime:FoundryLocalCore, qnn_execution_provider.cc:370 onnxruntime::QNNExecutionProvider::QNNExecutionProvider]
Unable to determine backend path from provider options. Using default.

Followed by node assignment warnings:

[W:onnxruntime:, session_state.cc:1316 onnxruntime::VerifyEachNodeIsAssignedToAnEp]
Some nodes were not assigned to the preferred execution providers which may or may not have a negative impact on performance.

Impact: Unclear whether the NPU is actually being utilised or if everything falls back to CPU. The warning offers no actionable guidance for developers.

Expected: Either suppress the warning when default fallback is the intended behaviour, or document which backend is actually selected.

[leestott]

Update (2026-03-11): As of CLI v0.8.117, QNN EP autoregistration succeeds (Successfully downloaded and registered the following EPs: QNNExecutionProvider). The backend path warning still appears during model loads but the NPU is functional.