Hi, I am trying to convert my model to tensorrt with int8. it works fine with batch size 1, but fails with batch size 4. The error message is
../builder/cudnnCalibrator.cpp:566: unsigned int nvinfer1::builder::nextPowerOf2(unsigned int): Assertion `q > p' failed.
I have no idea which part of my code is wrong. Any suggestion would be appreciated.
TensorRT version: 5.1.5