i'm having a problem where the application uses 50% of my cpu, but 7% of my gpu. cpu usage was 20% higher prior to converting it to onnx (and selecting the new file under model).
the only interesting bit from the output is the following:
2024-01-24 10:46:46.4732475 [W:onnxruntime:, session_state.cc:1171 onnxruntime::VerifyEachNodeIsAssignedToAnEp] Rerunning with verbose output on a non-minimal build will show node assignments.
i'm using the recommended settings by the bot and speed and all is fine.. but it would be nice to off load some of the cpu usage to the external gpu.
cpu: 7950X3D (32 threads)
gpu: 7900XTX
version: v1.5.3.17b onnxdirectML-cuda
windows 10 pro version 22H2 (OS Build 19045.3930)
