Giters

AlexeyAB / yolo2_light

Light version of convolutional neural network Yolo v3 & v2 for objects detection with a minimum of dependencies (INT8-inference, BIT1-XNOR-inference)

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Alexnet quant error

Enshall opened this issue · comments

commented

Hi AlexeyAB, @AlexeyAB
Thanks for your excelent work, I am now test it on Alexnet, I got the calibrated value on FC layer as same as conv, but didnt get the result expected.
the input_calibration of alexnet I got is 15.7778, 15.7815, 15.7815, 15.8131, 15.8131, 15.7815, 15.8131, 15.8131, 16, when bin_width=1.0 / 16 and max_bin= 4096 ...
Can you give me some advices?
Thank you a lot!

commented

ok， I make it. It just overflow.