Tensorrt 7.1 silently disable int8 calibration if network contains bool tensor

AI & Data Science Deep Learning (Training & Inference)
#1

OS: ubuntu 16.04
TensorRT: 7.1.3.4
CUDA: 10.2
CUDNN: 8.0

I’m working on network contains “equal” operator and where operators. If I add “equal” operator to network, then int8 calibration is disabled silently without any warning or error message produced. otherwise int8 calibration works well. I have provide int8 calibrator and set int8_mode to True.

#2

Hi, I am also confusing about these issues and I just there is also an equal operator.
For this pytorch :
keep = (hmax == heat).float()

So how do you fix the issues?