Description
In AI Hub, a user submitted a model with a DynamicQuantizeLinear node. When generating a QNN context binary, it fails with:
ERROR - Node image_QuantizeLinear: 'No translation registered for op type onnx_dynamicquantizelinear.'
However, when that .onnx model is run with the QNN EP, it just fails with a:
Non-zero status code returned while running QNN_9500122506797089069_52 node. Name:'QNNExecutionProvider_QNN_9500122506797089069_52_52' Status Message: QNN graph execute error. Error code: 6031 at location sequential_executor.cc:572 ExecuteKernel
The user doesn't have any information on how to fix the model. The QNN EP should fail and report an issue with the dynamic quantize linear node.
(For those that will work on this bug, please request permission from me and I will share the job so you have access to the following:)
Original job with 6031 error:
https://app.aihub.qualcomm.com/jobs/jpx69e19p
Compile job with sybnamic quantize linear error:
https://app.aihub.qualcomm.com/jobs/jgj2do6v5/