How to get the non-quantized model with TIDL-RT？

[yonger001]

With the 09.01 version SDK, we got the wrong model's outputs on TDA4 side. To deal with the above issue, we plan to compare the inference outputs between the PC and the TDA4 board. We configured the following quantization parameters to obtain the .bin model file, but the resulting model is still quantized(The size of the exported model is the same as during quantization.).

`
modelType = 2
numParamBits = 8
numFeatureBits = 8
quantizationStyle = 0
enableFloatPoint = 1
calibrationOption = 0
biasCalibrationIterations = 1

inputNetFile = "/data/****.onnx"
outputNetFile = "/data/model_net_0513.bin"
outputParamsFile = "/data/model_io_0513_"

inDataNorm = 1
inMean = 0 0 0
inScale = 0.00392156862745098 0.00392156862745098 0.00392156862745098
foldPreBnConv2D = 1
inWidth = 640
inHeight = 512
inNumChannels = 3
inFileFormat = 2
inDataFormat = 1

numFrames = 50
inData = "/data/calibration.txt"
postProcType = 0`

Therefore, to obtain a quantized-disabled model, what quantization parameters should we set instead?