0

I have trained AlexNet for my own data (gray level medical images with size 227*227). Now that I look into the caffemodel files, I see that except some lines in the beginning all the other values are ff7f or ffff like this:

enter image description here

additionally when I try to visualize the network weights by python, I get this waring: UserWarning: Warning: converting a masked element to nan. and nothing is shown in the figure!

Do you know what those ff7f or ffff values in my coffemodel are? why they are resulted? or any recommendation or any similar issues that have been disussed? please help me find out.

Cù Đức Hiếu
  • 5,569
  • 4
  • 27
  • 35
user6726469
  • 231
  • 1
  • 3
  • 14
  • Those are possibly little-endian encodings of "not a number" (NaN) values of 32-bit floats; `7fffffff` if you'd properly group and order it. –  Oct 16 '16 at 10:00
  • @Rhymoid would you please explain more? how can I group pr order it? – user6726469 Oct 16 '16 at 12:09
  • In your screenshot, the nybbles of the floating point values are grouped and ordered as `6745 2301 6745 2301`. You should read them in the order and grouping `01234567 01234567` if you want to interpret the data yourself. –  Oct 16 '16 at 13:31
  • Possible duplicate of [Common causes of nans during training](http://stackoverflow.com/questions/33962226/common-causes-of-nans-during-training) – Shai Oct 17 '16 at 19:15

0 Answers0