Part Number: TDA4VM
Hello experts,
My SDK version is processor-sdk-linux-sk-tda4vm-08_04_00.I trained the ssd model using the edgeai-mmdetection project.
I converted .pth to .onnx and .prototxt using the run_detection_export.sh.
Then, I want to use run_custom_pc.sh in the edgeai-benchmark-r8.4 project to generate the modelartifacts. Results show :[ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Unexpected input data type. Actual: (tensor(uint8)) , expected: (tensor(float))
The detailed steps of my operation are as follows:
The first step:
I trained the ssd model using the run_detection_train.sh in the edgeai-mmdetection project.It runs well and gets the .pt model.
The second step:
I convert the.pt model into .onnx and .prototxt files by using run_detection_export.sh.It works well.
The third step:
I want to use run_custom_pc.sh in the edgeai-benchmark-r8.4 project to generate the modelartifacts.
Is this because of the parameter setting problem in the second step?
Is it related to the content of the red box in the following figure? Where to set the flag quantize?
Thank you!
maiunlei
