Can not infer I16 input precision with I8 model presicion

TCE Options

TCE Open Date: 

Thursday, November 28, 2019 - 02:18

Can not infer I16 input precision with I8 model presicion


I am trying to use I16 input precision with I8 MNIST model precision.

Others input presicion(such as FP32, U8, U16) outputs correct value, but I16 input precision outputs incorrect value(I16 input presicion with FP32 model precision outputs correct value).

 

I got I8 model by

$python3 calibration_tool.py -sm -m <FP32 MNIST model>.xml --output-dir <output directory>.

 

I16 input precision with I8 model precision is not supported?

 

Thank you for help.

1 post / 0 new